![of my situation kept me awake the greater part of the night; and
very early the next morning I rose up from my mat in the tambu-
house to view, undisturbed, the interior region of the island. It was a
gloomy morning. Thin lines of mist were still encircling the loftier
summits or lingering in the valleys below. Here and there on the
crest of some distant hill a cluster of cocoa-nut palms marked the
home of a bush-tribe effectually isolated by deep intervening valleys
from the neighbouring tribes. I gazed upon a region which had for
ages worn the same aspect, inhabited by the same savage races,
the signs of whose existence played such an insignificant part in the
panorama laid out before me. Standing alone on this hill-top, I
reflected on the deeds of barbarity which these silent mountains
must have witnessed “in the days of other years,” deeds which are
only too frequent in our own day when the hand of every tribe is
against its neighbour, and when the butchery of some unsuspecting
hamlet too often supplies the captors with the materials for the
cannibal feast.
By the unusual success of their treachery and cunning—the two
weapons most essential to savage warfare in St. Christoval as well as
in the other islands—some chiefs have acquired a predominance
over the neighbouring villages, and their name inspires terror
throughout the island. Amongst them, I may mention Taki, the chief
of the large village of Wano on the north coast of this island. He has
obtained the double reputation of being a friend to the white man
and of being the most accomplished head-hunter in St. Christoval;
and, as may be readily imagined, the efforts of the Melanesian
Mission, by whom a station has been for many years established in
this village,[5] have been greatly retarded by the indifference of this
powerful chief. The resident teacher in the village was his own son,
who had been selected by Bishop Selwyn and had undergone the
usual training of teachers in Norfolk Island. I regret to write that he
greatly lapsed during our stay in the group, that he appears to have
accompanied his father on a head-hunting foray, and that he finally
met with an untimely fate, being so severely wounded by a shark
when fishing on the reef that he died a few hours afterwards. Taki,](https://image.slidesharecdn.com/5013-250422125545-4901ffca/85/Real-time-Speech-and-Music-Classification-by-Large-Audio-Feature-Space-Extraction-1st-Edition-Florian-Eyben-Auth-35-320.jpg)
![although not a Christian convert, was fond of displaying his
connection with the Mission. He showed me a certificate which he
received from Bishop Patteson in July, 1866; and in fact he is always
ready to do the honours of his village to the white man. Of his head-
hunting propensities, Captain Macdonald, an American trader
resident in Santa Anna, told us the following tale: Not long before
the arrival of H.M.S. “Lark” in the Solomon Islands, he was sailing
along the St. Christoval coast, when he met Taki in his war-canoe
proceeding on one of these expeditions. He endeavoured to place
hindrances in the chief’s way by telling him that he had native-
traders living at the different places on the coast where he intended
to land. But it was to no purpose. Taki saw the ruse, and taking it in
good part remarked to Captain Macdonald that he had apparently a
large number of natives trading for him. Waiting patiently until some
unfortunate bushmen ventured down on the reefs to fish, the Wano
chief surprised them, slaughtered many and carried the living and
the dead in triumph to his village. When Mr. Brenchley visited this
village in H.M.S. “Curacoa” in 1865, he saw evidence of a head-
hunting foray, in which probably Taki had taken part in his youthful
days. The skulls of 25 bushmen were observed hanging up under
the roof of the tambu-house, all showing the marks of the
tomahawk.[6] In our time, this chief conducted his forays less openly,
and I saw no evidence of his work in the tambu-houses of his
village.
[5] The Rev. J. Atkin was resident at Wano in 1871, shortly before he met
his death with Bishop Patteson in Santa Cruz.
[6] “Cruise of H.M.S. ‘Curacoa’” (p. 267); by J. L. Brenchley, M.A.
The practice of head-hunting, above referred to, prevails over a
large extent of the Solomon Group. The chiefs of New Georgia or
Rubiana extend their raids to Isabel, Florida, and Guadalcanar; and
thus perform voyages over a hundred miles in length. Within the
radius of these raids no native can be said to enjoy the security of
his own existence for a single day. In the villages of Rubiana may be
seen heaps of skulls testifying to the success of previous](https://image.slidesharecdn.com/5013-250422125545-4901ffca/85/Real-time-Speech-and-Music-Classification-by-Large-Audio-Feature-Space-Extraction-1st-Edition-Florian-Eyben-Auth-36-320.jpg)
![expeditions. Captain Cheyne, when visiting Simbo or Eddystone
Island in 1844, found that the natives had just returned from a
successful expedition, bringing with them 93 heads of men, women,
and children. In these expeditions, he says, they sometimes reached
as far as Murray Island which lies about 135 miles to the eastward.
[7] Their reputation, however, had extended yet further, since
D’Urville, who visited Thousand Ships Bay in 1838, tells us that the
Isabel natives knew the land of Simbo and pointed to the west to
indicate its direction.[8] The Rev. Dr. Codrington, in referring to these
head-hunting raids,[9] remarks that the people of the south-west
part of Isabel have suffered very much from attacks made on them
year after year by the inhabitants of the further coast of the same
island and of neighbouring islands, the object of these attacks being
to obtain heads, either for the honour of a dead or living chief or for
the inauguration of new canoes. He observes that a new war canoe
is not invested with due mana, i.e., supernatural power, until some
man has been killed by those on board her; and any unfortunate
voyagers are hunted down for the purpose on the first trip or
afterwards. The Rubiana natives are said to have introduced head-
hunting and human sacrifices into the neighbouring islands. They
carry off not only heads but living prisoners, whom they are believed
to keep, till on the death of a chief, or launching of a canoe, or some
great sacrifice, their lives are taken.
[7] “A Description of Islands in the Western Pacific Ocean” (p. 66), by
Andrew Cheyne, London, 1852.
[8] “Voyage au Pole Sud,” Paris, 1843; tom. v., p. 31.
[9] Journal of Anthropological Institute, vol. x., p. 261.
White men have sometimes been the victims of these head-hunting
expeditions. As is well known, Lieutenant Bower, of H.M.S. “Sandfly,”
met his death, together with the greater number of his boat’s crew,
on the islet of Mandoleana, in 1880, at the hands of a similar
expedition undertaken by the Florida natives. Kalikona, the most
influential chief of the Florida Islands, was freed from implication in
this tragedy mainly through the efforts of Bishop Selwyn, to whose](https://image.slidesharecdn.com/5013-250422125545-4901ffca/85/Real-time-Speech-and-Music-Classification-by-Large-Audio-Feature-Space-Extraction-1st-Edition-Florian-Eyben-Auth-37-320.jpg)
![any native who from his superior knowledge of English seemed to be
supplanting him in the intercourse with the ships that visited the
harbour. Even his right-hand man, who prided himself on his name
of Billy, experienced his wrath on one occasion in this manner. Like
other chiefs, Mule is grasping and covetous, shortcomings which are
rather those of the race than of the individual. Although of the chiefs
of Bougainville Straits I liked him the least, the contrast was rather
due to the exceptionally good estimate we had formed of his fellow
chiefs. The visits of H.M.S. “Lark” to this island have been the means
of removing the very bad reputation which the natives had
deservedly possessed: and I would especially invite the attention of
my readers to the history of this change in the attitude of these
natives towards the white man.
Captain C. H. Simpson, who visited this island in H.M.S. “Blanche” in
1872, described its people in his report to the Admiralty,[10] as being
“the most treacherous and blood-thirsty of any known savages;” and
the officers employed in making a sketch of the harbour had ample
evidence of their ferocity. About seven years before, the natives had
cut out a barque and had murdered her crew of 33 men. Previously
they had captured several boats of whalers visiting the islands, and
had massacred the crews. The Treasury natives were always very
reticent to us when we tried to learn something more of the fate of
the barque; but we learned little except that she was American, and
was named “Superior.” The captain, whose name the natives
pronounced “Hoody,” was carried away into the interior of the island
and killed, and the scene of his murder was once pointed out to
Lieutenant Oldham when crossing the island. As Captain Simpson
charges the natives with cannibalism, there can be little doubt of the
ultimate fate of the crew of the American barque. In the interval
between the occurrence of this event and the arrival of the
“Blanche,” no vessel had anchored in the harbour, the ships always
heaving-to off the north coast, where the natives resided when
Captain Simpson visited the island. Treasury retained its bad
reputation up to the date of our visit; and but few traders had much
knowledge of the place, as they generally gave the island a wide](https://image.slidesharecdn.com/5013-250422125545-4901ffca/85/Real-time-Speech-and-Music-Classification-by-Large-Audio-Feature-Space-Extraction-1st-Edition-Florian-Eyben-Auth-46-320.jpg)
![berth. We met but one man who spoke well of these natives, and he
was Captain Walsch of the trading schooner “Venture.” All others
gave them the worst of characters: and led me to believe that my
acquaintance with Treasury would not extend beyond the deck of
H.M.S. “Lark.” When Lieutenant Oldham first visited this island in
May, 1882, he had every reason to place but little confidence in the
natives; and in truth we all thought that the appearance and
behaviour of the natives justified the treacherous reputation which
they had obtained. Only two days were spent there, but no landing
was effected: the chief made no response to the invitations to visit
the ship; and we left the harbour without much feeling of regret. In
June of the following year we again visited this island; and if the
same procedure had been followed we should have been a very long
time in gaining the confidence of the natives. Lieutenant Oldham,
however, paid an official visit to the chief, accompanied by
Lieutenant Malan and myself. Mule and one of his sons returned the
visit within a couple of hours. Presents were exchanged; and the
foundation of mutual confidence was thus laid. The result may be
briefly stated. In a few days I was rambling all over the island,
usually accompanied by a lively gathering of men and boys. An
intimacy was established with the natives, which lasted until we
bade farewell to the group in the following year; and the return of
the “Lark” from her cruises was always a cause of rejoicing amongst
the natives. The men of the ship were known by name to most of
the people of the island: whilst Mr. Isabell, our leading-stoker, made
a deep impression upon them by his readiness to employ his
mechanical skill for their various wants, so much so that Mule
offered, if he would remain, to make him a chief with the usual
perquisite as to the number of his wives. For my own part, I reaped
the full benefit of our amicable relations with the natives; and for the
proof of this statement I must refer the reader to the remarks on my
intercourse with them, and to my observations on the geology,
botany, and other characteristics of the island.
[10] “Hydrographic Notices, Pacific Ocean,” 1856 to 1873 (p. 106).](https://image.slidesharecdn.com/5013-250422125545-4901ffca/85/Real-time-Speech-and-Music-Classification-by-Large-Audio-Feature-Space-Extraction-1st-Edition-Florian-Eyben-Auth-47-320.jpg)
![anchor his ships in 1768, being opposed by the hostility of the
natives. The boats, which had been sent in to find an anchorage,
were attacked by 150 men in ten canoes, who were only routed
after the second discharge of fire-arms. Two canoes were captured,
in one of which was found the jaw of a man half-broiled. The
number of shoals, and the irregularity of the currents prevented the
ships coming up to the anchorage before night fell; and Bougainville,
abandoning his design, continued his course through the Straits.[11]
The description which the French navigator gave of these natives in
1768, applies equally well to those of the present day. When H.M.S.
“Lark” revisited Choiseul Bay in October, 1884, not a single native
was seen; so that it would behove future visitors to be very cautious
in their dealings with these natives. Whilst off the coast north of this
bay, a fishing-party of half-a-dozen men came off to the ship from
the village of Kandelai; but they showed great suspicion of us. They
would not come alongside for some time; and when a present of
calico was flung to them at the end of a line, they were divided
amongst themselves whether to come and take it, some paddling
one way and some another. At length they took the present and
came alongside, but did not stay long, and soon paddled towards
the shore, their suspicions by no means allayed. What had happened
to cause this change of attitude, we could not learn. Evidently, the
good impression which we had left behind us a year before, had
borne no fruit. Probably, some inconsiderate action on the part of
the crew of a trading-vessel had undone our work.
[11] “Voyage autour du Monde,” 2nd edit. augm. vol. II., Paris, 1772.
The professional head-hunter of the eastern islands of the group
does not appear to be represented amongst the islands of
Bougainville Straits. Raids are occasionally made on the villages of
the adjoining Bougainville coast, but more, I believe, for the purpose
of procuring slaves, than from the mere desire of fighting. There is,
however, frequent friendly communication between the natives of
the islands of the Straits and those of certain Bougainville villages,
the former usually exchanging articles of trade for spears and](https://image.slidesharecdn.com/5013-250422125545-4901ffca/85/Real-time-Speech-and-Music-Classification-by-Large-Audio-Feature-Space-Extraction-1st-Edition-Florian-Eyben-Auth-50-320.jpg)
![who handled their canoes in an awkward manner or who stumbled
in their walk whilst accompanying me in my excursions. On one
occasion, when trying to obtain stone axes from the natives of Alu, I
was referred with a smile to the bushmen of the neighbouring island
of Bougainville, who still employ these tools. In the larger islands the
bush-tribes and the coast natives wage an unceasing warfare, in
which the latter are usually the aggressors and the victors—the
bushmen captured during these raids either affording materials for
the cannibal feast or being detained in servitude by their captors.
But there prevails in the group a recognized system of slave-traffic,
in which a human being becomes a marketable commodity—the
equivalent being represented in goods either of native or of foreign
manufacture. This custom which came under the notice of the
officers of Surville’s expedition, during their visit to Port Praslin in
Isabel, in 1769,[12] obtains under the same conditions at the present
time. These natives were in the habit of making voyages of ten and
twelve days’ duration with the object of exchanging men for “fine
cloths covered with designs,” articles which were manufactured by a
race of people much fairer than their own, who were in all
probability the inhabitants of Ontong Java.
[12] “Discoveries to the south-east of New Guinea,” by M. Fleurieu, p. 143,
Eng. edit.
The servitude to which the victims of this traffic are doomed is not
usually an arduous one. But there is one grave contingency attached
to his thraldom which must be always before the mind of the
captive, however lightly his chains of service may lie upon him.
When a head is required to satisfy the offended honour of a
neighbouring chief, or when a life has to be sacrificed on the
completion of a tambu-house or at the launching of a new war-
canoe, the victim chosen is usually the man who is not a free-born
native of the village. He may have been bought as a child and have
lived amongst them from his boyhood up, a slave only in name, and
enjoying all the rights of his fellow natives. But no feelings of
compassion can save him from his doom; and the only consideration
which he receives at the hands of those with whom he may have](https://image.slidesharecdn.com/5013-250422125545-4901ffca/85/Real-time-Speech-and-Music-Classification-by-Large-Audio-Feature-Space-Extraction-1st-Edition-Florian-Eyben-Auth-57-320.jpg)
![I have previously referred to the existence of bushmen on some of
the smaller islands. In the interior of Treasury there are a few
hamlets containing each two or three families of bushmen, who live
quite apart from the other natives of the island. On more than one
occasion I experienced the hospitality of these bush families, who in
matters of dress are even less observant than the harbour natives.
They are probably the remnants of the original bushmen who
occupied this island. Over our pipes, I used frequently to converse
with the natives on the subject of the past history of their island;
and I gleaned from them that the enterprising race at present
dominant in the Bougainville Straits came originally from the islands
immediately to the eastward, using Treasury as a stepping-stone to
the Shortlands and Faro, and ousting or exterminating the bushmen
they found in the possession of these islands.
I will turn for a moment to the subject of slavery in the eastern
islands of the group. In Ugi it is the practice of infanticide which has
given rise to a slave-commerce regularly conducted with the natives
of the interior of St. Christoval. Three-fourths of the men of this
island were originally bought as youths to supply the place of the
natural offspring killed in infancy. But such natives when they attain
manhood virtually acquire their independence, and their original
purchaser has but little control over them. On page 42, I have made
further reference to this subject.
Connected in the manner above shown with the subject of slavery is
the practice of cannibalism. The completion of a new tambu-house is
frequently celebrated among the St. Christoval natives by a cannibal
feast. Residents in that part of the group tell me that if the victim is
not procured in a raid amongst the neighbouring tribes of the
interior, some man is usually selected from those men in the village
who were originally purchased by the chief. The doomed man is not
enlightened as to the fate which awaits him, and may, perhaps, have
been engaged in the erection of the very building at the completion
of which his life is forfeited. The late Mr. Louis Nixon,[13] one of
those traders whose name should not be forgotten amongst the](https://image.slidesharecdn.com/5013-250422125545-4901ffca/85/Real-time-Speech-and-Music-Classification-by-Large-Audio-Feature-Space-Extraction-1st-Edition-Florian-Eyben-Auth-59-320.jpg)
![pioneers who, in working for themselves, have worked indirectly for
the good of their successors in the Solomon Group, once recounted
to me a tragical incident of this kind on the island of Guadalcanar, of
which he was an unwilling spectator. Whilst looking out of the
window of his house one afternoon, he observed a native walk up to
another standing close to the window and engage him in
conversation. A man then stole up unperceived, and raising his
heavy club above his head, struck the intended victim lifeless to the
ground. Knowing too well the nature and purpose of the deed, Mr.
Nixon turned away quite sickened by the sight.
[13] Mr. Nixon died at Santa Anna in the end of 1882.
The natives of the small island of Santa Anna enjoy the reputation of
being abstainers from human flesh: but, inasmuch, as Mai the war-
chief has acquired a considerable fortune, in a native’s point of view,
by following the profitable calling of purveyor of human flesh to the
man-eaters of the adjacent coasts of St. Christoval—a trade in which
he is ably assisted by those who accompany him on his foraging
expeditions—we can hardly preserve this nice distinction between
the parts taken by the contractor and his customers in this
extraordinary traffic. I learned from Captain Macdonald that in their
abstinence from human flesh, the Santa Anna natives are not
actuated by any dislike of anthropophagy in itself; but that the
custom has fallen into abeyance since the chief laid the tambu-ban
on human flesh several years ago, on account of a severe epidemic
of sickness having followed a cannibal feast. On one occasion
through the instrumentality of this resident, Lieutenant Oldham had
the satisfaction of rescuing two St. Christoval natives whom Mai was
carefully keeping in anticipation of the wants of the man-eaters of
Cape Surville. As the result of an interview held with this chief, the
two prisoners were sent on board the “Lark;” but Mai gave them up
with a very bad grace, protesting that he was being robbed of his
own property. It is difficult to speculate on the reflections of the
victim as he lives on from day to day in constant expectation of his
fate. I am told that there is a faint gleam of tender feeling shown in](https://image.slidesharecdn.com/5013-250422125545-4901ffca/85/Real-time-Speech-and-Music-Classification-by-Large-Audio-Feature-Space-Extraction-1st-Edition-Florian-Eyben-Auth-60-320.jpg)
![the case of a man who, by long residence in the village, has almost
come to be looked upon as one of themselves. He is allowed to
remain in ignorance of the dreaded moment until the last: and,
perhaps, he may be standing on the beach assisting in the launching
of the very canoe in which he is destined to take his final journey,
when suddenly he is laid hold of, and in a few moments more he is
being ferried across to the man-eaters of the opposite coast. All
persons whom I have met that have had a lengthened experience of
the St. Christoval natives confirm these cannibal practices. They may
sometimes be observed with all the horrible preliminaries which have
been described in the cases of other Pacific groups; whilst, on the
other hand, it may be the habit to purchase and partake of human
flesh as an extra dainty in the daily fare.
Captain Redlich, master of the schooner “Franz,” who visited Makira
on the south side of St. Christoval in 1872, states that he found a
dead body in a war-canoe dressed and cooked whole. He was
informed by Mr. Perry, a resident, that he had seen as many as
twenty bodies lying on the beach dressed and cooked.[14] In 1865,
Mr. Brenchley noticed at Wano, on the north coast of this island, the
skulls of twenty-five bushmen hanging up under the roof of the
tambu-house, all of which showed the effects of the tomahawk and
all had been eaten.[15] At the present time it is not an easy matter
for any person not resident in the group to obtain ocular evidence of
cannibalism, since the natives have become aware of the white
man’s aversion to the custom. I have, however, frequently seen the
arm and leg bones of the victim consumed at the opening of a new
tambu-house, as they are usually hung up over the entrance or in
some other part of the building. The natives, however, are generally
reluctant to talk much about these matters; and I believe the
residents, in such matters, prefer to trust more to the testimony of
their own eyes than to the statements of the natives.
[14] Journal of the Royal Geographical Society for 1874 (vol. 44), p. 31.
[15] “Cruise of H.M.S. ‘Curaçoa,’” by J. L. Brenchley.](https://image.slidesharecdn.com/5013-250422125545-4901ffca/85/Real-time-Speech-and-Music-Classification-by-Large-Audio-Feature-Space-Extraction-1st-Edition-Florian-Eyben-Auth-61-320.jpg)
Real time Speech and Music Classification by Large Audio Feature Space Extraction 1st Edition Florian Eyben (Auth.)
- 1. Real time Speech and Music Classification by Large Audio Feature Space Extraction 1st Edition Florian Eyben (Auth.) download https://textbookfull.com/product/real-time-speech-and-music- classification-by-large-audio-feature-space-extraction-1st- edition-florian-eyben-auth/ Download more ebook from https://textbookfull.com
- 2. We believe these products will be a great fit for you. Click the link to download now, or visit textbookfull.com to discover even more! Image Color Feature Extraction Techniques Fundamentals and Applications Jyotismita Chaki https://textbookfull.com/product/image-color-feature-extraction- techniques-fundamentals-and-applications-jyotismita-chaki/ Feature extraction and image processing for computer vision Fourth Edition Aguado https://textbookfull.com/product/feature-extraction-and-image- processing-for-computer-vision-fourth-edition-aguado/ Arduino Music and Audio Projects Cook https://textbookfull.com/product/arduino-music-and-audio- projects-cook/ Audio And Speech Processing With MATLAB Paul R. Hill https://textbookfull.com/product/audio-and-speech-processing- with-matlab-paul-r-hill/
- 3. Feature Extraction and Image Processing for Computer Vision 4th Edition Mark Nixon https://textbookfull.com/product/feature-extraction-and-image- processing-for-computer-vision-4th-edition-mark-nixon/ Artificial Intelligence in Label free Microscopy Biological Cell Classification by Time Stretch 1st Edition Ata Mahjoubfar https://textbookfull.com/product/artificial-intelligence-in- label-free-microscopy-biological-cell-classification-by-time- stretch-1st-edition-ata-mahjoubfar/ Building Machine Learning Systems with a Feature Store: Batch, Real-Time, and LLM Systems (Early Release) Jim Dowling https://textbookfull.com/product/building-machine-learning- systems-with-a-feature-store-batch-real-time-and-llm-systems- early-release-jim-dowling/ Speech and Audio Processing A MATLAB based Approach 1st Edition Ian Vince Mcloughlin https://textbookfull.com/product/speech-and-audio-processing-a- matlab-based-approach-1st-edition-ian-vince-mcloughlin/ Mastering Kafka Streams and ksqlDB: Building Real-Time Data Systems by Example 1st Edition Mitch Seymour https://textbookfull.com/product/mastering-kafka-streams-and- ksqldb-building-real-time-data-systems-by-example-1st-edition- mitch-seymour/
- 4. SpringerTheses Recognizing Outstanding Ph.D. Research Real-timeSpeech andMusic Classificationby LargeAudioFeature SpaceExtraction Florian Eyben
- 5. Springer Theses Recognizing Outstanding Ph.D. Research
- 6. Aims and Scope The series “Springer Theses” brings together a selection of the very best Ph.D. theses from around the world and across the physical sciences. Nominated and endorsed by two recognized specialists, each published volume has been selected for its scientific excellence and the high impact of its contents for the pertinent field of research. For greater accessibility to non-specialists, the published versions include an extended introduction, as well as a foreword by the student’s supervisor explaining the special relevance of the work for the field. As a whole, the series will provide a valuable resource both for newcomers to the research fields described, and for other scientists seeking detailed background information on special questions. Finally, it provides an accredited documentation of the valuable contributions made by today’s younger generation of scientists. Theses are accepted into the series by invited nomination only and must fulfill all of the following criteria • They must be written in good English. • The topic should fall within the confines of Chemistry, Physics, Earth Sciences, Engineering and related interdisciplinary fields such as Materials, Nanoscience, Chemical Engineering, Complex Systems and Biophysics. • The work reported in the thesis must represent a significant scientific advance. • If the thesis includes previously published material, permission to reproduce this must be gained from the respective copyright holder. • They must have been examined and passed during the 12 months prior to nomination. • Each thesis should include a foreword by the supervisor outlining the signifi- cance of its content. • The theses should have a clearly defined structure including an introduction accessible to scientists not expert in that particular field. More information about this series at http://www.springer.com/series/8790
- 7. Florian Eyben Real-time Speech and Music Classification by Large Audio Feature Space Extraction Doctoral Thesis accepted by the Technische Universität München, Germany 123
- 8. Author Dr. Florian Eyben Institute for Human-Machine Communication (MMK) Technische Universität München Munich, Germany Supervisors Prof. Björn Schuller Department of Computing Imperial College London, UK and Chair of Complex and Intelligent Systems University of Passau Passau, Germany Prof. Werner Hemmert Bio-Inspired Information Processing Institute of Medical Engineering (IMETUM) Technische Universität München Munich, Germany ISSN 2190-5053 ISSN 2190-5061 (electronic) Springer Theses ISBN 978-3-319-27298-6 ISBN 978-3-319-27299-3 (eBook) DOI 10.1007/978-3-319-27299-3 Library of Congress Control Number: 2015957094 © Springer International Publishing Switzerland 2016 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. Printed on acid-free paper This Springer imprint is published by SpringerNature The registered company is Springer International Publishing AG Switzerland
- 9. Parts of this thesis have been published in the following articles: Journals • F. Eyben, M.Wöllmer, A. Graves, B. Schuller, E. Douglas-Cowie, and R. Cowie. Online emotion recognition in a 3-D activation-valence-time continuum using acoustic and linguistic cues. Journal on Multimodal User Interfaces (JMUI), 3(1–2):7–19, March 2010. doi:10.1007/s12193-009-0032-6 • F. Eyben, M.Wöllmer, and B. Schuller. A Multi-Task Approach to Continuous Five- Dimensional Affect Sensing in Natural Speech. ACM Transactions on Interactive Intelligent Systems, Special Issue on Affective Interaction in Natural Environments, 2(1), March 2012. Article No. 6, 29 pages • F. Eyben, A. Batliner, and B. Schuller. Towards a standard set of acoustic features for the processing of emotion in speech. Proceedings of Meetings on Acoustics (POMA), 9(1):1–12, July 2012 • F. Eyben, K. Scherer, B. Schuller, J. Sundberg, E. André, C. Busso, L. Devillers, J. Epps, P. Laukka, S. Narayanan, and K. Truong. The Geneva Minimalistic Acoustic Parameter Set (GeMAPS) for Voice Research and Affective Computing. IEEE Transactions on Affective Computing, 2015. 14 pages, in press • F. Eyben, G. L. Salomão, J. Sundberg, K. R. Scherer, and B. Schuller. Emotion in The Singing Voice—A Deeper Look at Acoustic Features in the Light of Automatic Classification. EURASIP Journal on Audio, Speech, and Music Processing, Special Issue on Scalable Audio-Content Analysis, 2015, 14 pages, in press Conferences • B. Schuller, B. Vlasenko, F. Eyben, G. Rigoll, and A. Wendemuth. Acoustic emotion recognition: A benchmark comparison of performances. In Proc. of the IEEE Workshop on Automatic Speech Recognition and Understanding (ASRU) 2009, pages 552–557, Merano, Italy, December 2009. IEEE • F. Eyben, M. Wöllmer, and B. Schuller. openEAR—Introducing the Munich Open-Source Emotion and Affect Recognition Toolkit. In Proc. of the 3rd International Conference on Affective Computing and Intelligent Interaction (ACII 2009), volume I, pages 576–581, Amsterdam, The Netherlands, September 2009. IEEE • F. Eyben, M. Wöllmer, and B. Schuller. openSMILE—The Munich Versatile and Fast Open-Source Audio Feature Extractor. In Proc. of ACM Multimedia 2010, pages 1459–1462, Florence, Italy, 2010. ACM
- 10. • F. Eyben, F.Weninger, F. Gross, and B. Schuller. Recent developments in openSMILE, the munich open-source multimedia feature extractor. In Proc. of ACM Multimedia 2013, pages 835–838, Barcelona, Spain, 2013. ACM • F. Eyben, B. Schuller, and G. Rigoll. Improving Generalisation and Robustness of Acoustic Affect Recognition. In L.-P. Morency, D. Bohus, H. K. Aghajan, J. Cassell, A. Nijholt, and J. Epps, editors, Proc. of the 14th ACM International Conference on Multimodal Interaction (ICMI) 2012, pages 517–522, Santa Monica, CA, USA, October 2012. ACM • F. Eyben, F. Weninger, S. Squartini, and B. Schuller. Real-life Voice Activity Detection with LSTM Recurrent Neural Networks and an Application to Hollywood Movies. In Proc. of ICASSP 2013, pages 483–487, Vancouver, Canada, May 2013. IEEE • F. Eyben, F. Weninger, and B. Schuller. Affect recognition in real-life acoustic conditions—A new perspective on feature selection. In Proc. of INTERSPEECH 2013, pages 2044–2048, Lyon, France, August 2013. ISCA
- 11. Supervisor’s Foreword It is an honour for me to introduce Dr. Eyben’s outstanding doctoral thesis work accepted for publication in Springer Theses with this Foreword. Dr. Eyben was one of the first two Ph.D. students who joined my Machine Intelligence and Signal Processing research group at the Insititute for Human-Machine Communication of the Technische Universität München (TUM) in 2008. He worked on the award-winning EU-FP7 research project SEMAINE for 3 years, where he devel- oped fundamental methods for automatic, real-time speech emotion recognition and non-verbal vocal analysis in the context of emotional sensitive virtual agent char- acters. During this period he created and maintained the openSMILE software toolkit for acoustic feature extraction and paralinguistic speech analysis. This toolkit and the methods implemented by it, laid the foundation of his ground-breaking thesis work. It was awarded twice at the ACM International Conference on Multimedia (2010 and 2013) and has been used to provide com- parative baseline evaluations in several international, renowned research competi- tions in the field. During his time at my research group, he was a big help for supporting and advising master students and new Ph.D. students as the group quickly grew in size. He constructively advised other team members on their research and collaborated with them in many studies and publications, contributing largely to the success and the international standing of the whole group. Besides his project work, in his spare time he volunteered to help me with lecture preparation and to hold a tutorial course for the Pattern Recognition lecture offered at our institute. For the excellent tutorial, which was packed with hands-on examples and online demonstrations for most of the topics covered, he was awarded the best lecture award from the student association of the faculty for electrical engineering. Automatic paralinguistic speech analysis is a young research field pioneered by my group since 2010. It is the continuation and generalisation of automatic speech emotion recognition, which has been an active research field for almost two dec- ades. The goal of speech emotion recognition is to automatically provide an vii
- 12. estimate of a speaker’s emotional state from the acoustic properties of her/his voice. Psychological studies have identified numerous acoustic properties which are correlated to emotion. However, these properties are high-level descriptions of acoustic vocal qualities and cannot be robustly extracted from speech recordings by standard signal processing algorithms. Thus, inspired by my earlier work on acoustic feature brute-forcing for speech emotion recognition, Dr. Eyben adopted and extended this approach for his thesis work. The fundamental idea of this method is to compute a very high number of acoustic descriptors from a speech signal, regardless of their theoretical significance or correlation with a specific task, and then apply machine learning methods on large data sets to create robust models for identification of paralinguistic speech information, such as a speaker’s emotion, age, gender, personality, alcohol intoxication level, sleepiness, depression, or cer- tain voice pathologies. Dr. Eyben conducted original research work on a previously unseen large amount of acoustic descriptors which he implemented in his open-source software toolkit openSMILE. This toolkit is the first of its kind which is capable of extracting all descriptors incrementally, in real-time. Further, he was the first to publish an open-source emotion recognition toolkit (openEAR), based on openSMILE. All of this have opened completely new possibilities for research teams across the world, such as integrating out-of-the-box emotion recognition into their projects, rapidly creating interactive prototypes from their research, and—most notably—investigating automatic recognition of paralinguistic information besides emotion. Renowned research competitions which provide comparative evaluations for several automatic recognition tasks in the field were organised by my group annually at INTERSPEECH and other highly recognised venues. The baseline acoustic descriptors and baseline results were created based on Dr. Eyben’s thesis work, which thereby has set a widely acknowledged standard for international comparison—a high standard even—as these baseline results were often hard to beat despite the large number of participating teams which included top research teams from all over the world. Recently, audEERING, a spin-off company was founded by me, Dr. Eyben, and colleagues. Having bought the intellectual property rights to the software and methods developed by Dr. Eyben at TUM, audEERING is now successfully marketing the emotion recognition and non-verbal music and speech analytics technology for applications such as call and contact centre quality monitoring, marketing research, brand testing, human health and safety, as well as security and entertainment products. Munich Prof. Björn Schuller February 2015 viii Supervisor’s Foreword
- 13. Abstract Automatic classification of speech and music has become an important topic with the advent of speech technologies in devices of our daily lives, such as smartphones and smartwatches. While for automatic speech recognition, commercial technolo- gies with good accuracies are available to buy, the classification of music and paralinguistic information beyond the textual content, such as mood, emotion, voice quality, or personality, is a very young field, but possibly the next technological milestone for man–machine interfaces. This thesis advances the state of the art in the area by defining standard acoustic feature sets for real-time speech and music analysis and by proposing solutions for real-world problems: a multi-condition learning approach to increase noise robustness, noise robust incremental segmentation of the input audio stream based on a novel, context-aware, and data-driven voice activity detector, and a method for fully (time and value) continuous affect regression tasks are introduced. Standard acoustic feature sets were defined and evaluated throughout a series of international research challenges. Further, a framework for incremental and real-time acoustic feature extraction is proposed, implemented, and published as an open-source toolkit (openSMILE). The toolkit implements all of the proposed baseline acoustic feature sets and has been widely accepted by the community—the publications introducing the toolkit have been cited over 400 times. The proposed acoustic feature sets are systematically evaluated on 13 databases containing speech affect and music style classification tasks. Experiments over a wide range of conditions are performed, i. e. training instance balancing, feature value normalisation, and various classifier parameters. Also, the proposed methods for real-time, noise robust, and incremental input segmentation and noise robust multi-condition classification are extensively evaluated on several databases. Finally, fully continuous (time and value), automatic recognition of affect in five dimensions with long short-term memory recurrent neural networks is evaluated on a database of natural and spontaneous affect expressions (SEMAINE). The superior performance of the proposed large feature sets over smaller sets was shown suc- cessfully for a multitude of tasks. ix
- 14. All in all, this thesis is a significant contribution to the field of speech and music analysis and hopefully expedites the process of bringing real-world speech and music analysis applications, such as robust emotion or music mood recognition, a bit closer to daily use. x Abstract
- 15. Acknowledgments This thesis is a result of my research work conducted at the Institute for Human-Machine Communication at the Technische Universität München, Germany. I would like to thank my supervisor Prof. Björn Schuller for providing the opportunity and resources to write my thesis in his Machine Intelligence and Signal Processing group at the Institute. He was an excellent mentor and supervisor, leaving me enough freedom to conduct research in the areas of my interest but giving me enough guidance to help me focus my research on important issues, learn how to write scientific papers, and publish my work successfully. I enjoyed the great collaborative atmosphere while working in his team and I am very grateful to have been introduced to and get to work with many high-level international col- laborators through him and his projects. I further want to thank him for providing me the opportunity to do two internships abroad during the time of my thesis. I am grateful to Prof. Gerhard Rigoll, who—as the head of the Institute—made my stay at the institute possible in the first place and constantly supported my research and the research of the group. I would like to thank, both him and Prof. Schuller, for their support which enabled me to attend many international confer- ences and project meetings. I would further like to thank my second examiner Prof. Werner Hemmert for his time and effort dedicated to my thesis, as well as Prof. Wolfgang Utschick for hosting my defence at his institute. Special thanks go to all my colleagues who have worked with me and who have always been available for inspiring and challenging discussions. I especially want to thank my co-authors, Felix Weninger, with whom I had long and inspiring dis- cussions and have developed many ideas and publications together, and Martin Wöllmer, who extensively collaborated with me and contributed to the success of the SEMAINE project. Further, I want to thank my colleague Erik Marchi for his testing and coding work on the openSMILE GUI parts and for valuable in-office discussions, as well as Jürgen Geiger and Fabien Ringeval, for their support and discussions. xi
- 16. For his great coordination of the SEMAINE project, his integration efforts, and the wonderful coder-camps, I would like to thank Marc Schröder. For a fruitful collaboration in the SEMAINE project and for helping me find accommodation during my internship in London, I would like to thank Michel Valstar. For excellent discussions on acoustic parameters, for providing the GEMEP corpus, for his lead efforts in proposing the Geneva Minimalistic Parameter Set, and for his support for the large scale parameter set evaluations, I would like to express special thanks to Prof. Klaus Scherer. I also want to thank my master students Christoph Kozielski, Benedikt Gollan, Marcel Knapp, and Bernd Huber, for their contributions of ideas and for testing and improving openSMILE throughout their thesis work. Also, I want to thank all of the many researchers around the world who actively use openSMILE, give feedback, and thereby contribute to the success of the standard feature sets and the toolkit. Most of all, for their encouraging support, love, and care, I would like to thank my wife Sarah and my parents. Munich Florian Eyben June 2014 The research leading to these results has received funding from the European Community’s Seventh Framework Programme (FP7/2007–2013) under grant agreements No. 211486 (SEMAINE) and No. 289021 (ASC-Inclusion). The research was further supported by an ERC Advanced Grant in the European Community’s Seventh Framework Programme under grant agreement 230331-PROPEREMO (Production and perception of emotion: an affective sci- ences approach) awarded to Prof. Klaus Scherer and hosted by the University of Geneva. xii Acknowledgments
- 17. Contents 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1 Acoustic Analysis of Speech and Music . . . . . . . . . . . . . . . . . . 2 1.2 Deficiencies of the State-of-the-Art. . . . . . . . . . . . . . . . . . . . . . 2 1.3 Aims of This Thesis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.3.1 Real-time Analysis Framework . . . . . . . . . . . . . . . . . . 3 1.3.2 Baseline Feature Sets . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.3.3 Real-World Robustness. . . . . . . . . . . . . . . . . . . . . . . . 4 1.3.4 Large-Scale Evaluation . . . . . . . . . . . . . . . . . . . . . . . . 5 1.4 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2 Acoustic Features and Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.1 Basics of Signal Processing. . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.1.1 Signal Representation . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.1.2 Frequency Domain . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2.1.3 Short-Time Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . 13 2.1.4 Pre-processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 2.2 Acoustic Low-Level Descriptors . . . . . . . . . . . . . . . . . . . . . . . 19 2.2.1 Time Domain Descriptors . . . . . . . . . . . . . . . . . . . . . . 20 2.2.2 Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 2.2.3 Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 2.2.4 Spectral Descriptors . . . . . . . . . . . . . . . . . . . . . . . . . . 35 2.2.5 Autocorrelation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 2.2.6 Cepstrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 2.2.7 Linear Prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 2.2.8 Formants. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 2.2.9 Perceptual Linear Prediction . . . . . . . . . . . . . . . . . . . . 55 2.2.10 Cepstral Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 2.2.11 Pitch. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 2.2.12 F0 Harmonics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 2.2.13 Voice Quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 xiii
- 18. 2.2.14 Tonal Features. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 2.2.15 Non-linear Vocal Tract Model Features. . . . . . . . . . . . . 80 2.3 Derived Features and Post-processing of Low-Level Descriptors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 2.3.1 Differences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 2.3.2 Delta Regression Coefficients. . . . . . . . . . . . . . . . . . . . 83 2.3.3 Higher Order Delta Regression Coefficients and Differences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 2.3.4 Temporal Smoothing . . . . . . . . . . . . . . . . . . . . . . . . . 84 2.4 Supra-Segmental Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 2.4.1 Stacking of Low-Level Descriptors. . . . . . . . . . . . . . . . 85 2.4.2 Statistical Functionals . . . . . . . . . . . . . . . . . . . . . . . . . 86 2.4.3 Modulation Functionals. . . . . . . . . . . . . . . . . . . . . . . . 103 2.5 Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 2.5.1 Static Modelling with Support Vector Machines. . . . . . . 107 2.5.2 Dynamic Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . 109 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 3 Standard Baseline Feature Sets. . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 3.1 INTERSPEECH 2009 Emotion Challenge Set . . . . . . . . . . . . . . 124 3.2 INTERSPEECH 2010 Paralinguistics Challenge Set . . . . . . . . . . 124 3.3 INTERSPEECH 2011 Speaker State Challenge Set . . . . . . . . . . 126 3.4 INTERSPEECH 2012 Speaker Trait Challenge Set . . . . . . . . . . 127 3.5 INTERSPEECH 2013 ComParE Set. . . . . . . . . . . . . . . . . . . . . 128 3.6 INTERSPEECH 2014 ComParE Set. . . . . . . . . . . . . . . . . . . . . 130 3.7 Audio-Visual Emotion Challenge Sets . . . . . . . . . . . . . . . . . . . 130 3.8 Geneva Minimalistic Acoustic Parameter Set. . . . . . . . . . . . . . . 131 3.9 Music Genre Sets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 3.10 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 4 Real-time Incremental Processing . . . . . . . . . . . . . . . . . . . . . . . . . . 139 4.1 Segmentation Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 4.1.1 On-Line Segmentation . . . . . . . . . . . . . . . . . . . . . . . . 141 4.1.2 Incremental Segmentation . . . . . . . . . . . . . . . . . . . . . . 142 4.2 Feature Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143 4.3 Architecture of the openSMILE Framework . . . . . . . . . . . . . . . 144 4.3.1 Incremental Processing . . . . . . . . . . . . . . . . . . . . . . . . 147 4.3.2 Smile Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151 4.4 Fully Continuous Speech Emotion Recognition . . . . . . . . . . . . . 151 4.4.1 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152 4.4.2 Proposed Continuous Modelling Approach . . . . . . . . . . 153 4.4.3 Acoustic Features. . . . . . . . . . . . . . . . . . . . . . . . . . . . 156 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157 xiv Contents
- 19. 5 Real-Life Robustness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 5.1 Voice Activity Detection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164 5.1.1 Related VAD Approaches . . . . . . . . . . . . . . . . . . . . . . 164 5.1.2 Proposed VAD Based on LSTM-RNNs . . . . . . . . . . . . 166 5.1.3 Benchmarking of the Proposed Approach . . . . . . . . . . . 167 5.2 Feature Normalisation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171 5.2.1 Normalisation of Low-Level Descriptors . . . . . . . . . . . . 172 5.2.2 Normalisation of Supra-Segmental Features. . . . . . . . . . 173 5.2.3 Incremental Normalisation. . . . . . . . . . . . . . . . . . . . . . 174 5.3 Noise Robustness. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 5.3.1 Synthesis of Noisy and Reverberated Data . . . . . . . . . . 177 5.3.2 Acoustic Feature Analysis and Selection . . . . . . . . . . . . 179 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180 6 Evaluation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185 6.1 Speech and Music Databases. . . . . . . . . . . . . . . . . . . . . . . . . . 185 6.1.1 Airplane Behaviour Corpus (ABC). . . . . . . . . . . . . . . . 186 6.1.2 FAU-AIBO Database (AIBO) . . . . . . . . . . . . . . . . . . . 187 6.1.3 TUM Audiovisual Interest Corpus (AVIC) . . . . . . . . . . 188 6.1.4 Danish Emotional Speech Database (DES) . . . . . . . . . . 189 6.1.5 Berlin Emotional Speech Database (EMO-DB) . . . . . . . 189 6.1.6 eNTERFACE’05 Database . . . . . . . . . . . . . . . . . . . . . 190 6.1.7 Geneva Multimodal Emotion Portrayals (GEMEP) . . . . . 190 6.1.8 Belfast Sensitive Artificial Listener Database (SAL) . . . . 191 6.1.9 SEMAINE Database. . . . . . . . . . . . . . . . . . . . . . . . . . 192 6.1.10 Geneva Singing Voice Emotion (GeSiE) Database . . . . . 198 6.1.11 Speech Under Simulated and Actual Stress (SUSAS). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199 6.1.12 Vera-Am-Mittag (VAM) . . . . . . . . . . . . . . . . . . . . . . . 200 6.1.13 Ballroom Dance-Style Database (BRD). . . . . . . . . . . . . 200 6.1.14 Genre Discrimination Database (GeDiDB). . . . . . . . . . . 201 6.2 Noise Robust Affective Speech Classification . . . . . . . . . . . . . . 202 6.2.1 Analysis of Acoustic Features . . . . . . . . . . . . . . . . . . . 202 6.2.2 Classification Performance. . . . . . . . . . . . . . . . . . . . . . 203 6.3 Evaluation of the Baseline Feature Sets . . . . . . . . . . . . . . . . . . 207 6.3.1 Mapping of Emotions . . . . . . . . . . . . . . . . . . . . . . . . . 207 6.3.2 Evaluation Method. . . . . . . . . . . . . . . . . . . . . . . . . . . 208 6.3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211 6.4 Continuous Dimensional Affect Recognition . . . . . . . . . . . . . . . 225 6.4.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . 226 6.4.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234 Contents xv
- 20. 7 Discussion and Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237 7.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237 7.2 Achievements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239 7.3 Future Work and Concluding Remarks . . . . . . . . . . . . . . . . . . . 241 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242 Appendix A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247 Appendix B: Mel-Frequency Filterbank Parameters. . . . . . . . . . . . . . . 295 Curriculum Vitae—Florian Eyben . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297 xvi Contents
- 21. Acronyms ABC Airplane Behaviour Corpus ACF Autocorrelation Function AR Autoregressive ARFF Attribute Relation Feature Format AR-GARCH Autoregressive-Generalised Autoregressive Conditional Heteroskedasticity ASR Automatic Speech Recognition AUC Area Under (ROC) Curve AVEC Audio-Visual Emotion Challenge AVIC Audio-Visual Interest Corpus BLSTM Bidirectional Long Short-Term Memory BLSTM-RNN Bidirectional Long Short-Term Memory Recurrent Neural Network BPM Beats per Minute BPTT Backpropagation Through Time BRD BallRoom Dance-style CC (Pearson) Correlation Coefficient CEC Constant Error Carousel CENS CHROMA Energy-distribution Normalised Statistics CEP Cepstrum CFS Correlation-based Feature-subset Selection ComParE Computational Paralinguistics ChallengE CV Coefficient of Variation dB Decibel DC Direct current DCT Discrete Cosine Transformation DCT-II Discrete Cosine Transformation Type-II DES Danish Emotional Speech DFT Discrete Fourier Transformation DSP Digital Signal Processor EER Equal Error Rate xvii
- 22. eGeMAPS Extended Geneva Minimalistic Acoustic Parameter Set EMI Electromagnetic Interference EMO-DB Berlin Emotional Speech Database EOI End-of-Input FFNN Feed-Forward Neural Network FFT Fast Fourier Transformation FIR Finite Impulse Response FNR False Negative Rate FPGA Field Programmable Gate Array FPR False Positive Rate FT Fourier Transformation GeMAPS Geneva Minimalistic Acoustic Parameter Set GEMEP Geneva Multimodal Emotion Portrayals GMM Gaussian Mixture Model HMM Hidden Markov Model HNR Harmonics-to-Noise Ratio HPCP Harmonic Pitch Class Profiles HTK Hidden Markov Toolkit (Young et al. 2006) Hz Hertz IIR Infinite Impulse Response IQR Inter-Quartile Range IS09 INTERSPEECH 2009 Emotion Challenge IS10 INTERSPEECH 2010 Paralinguistics Challenge IS11 INTERSPEECH 2011 Speaker State Challenge IS12 INTERSPEECH 2012 Speaker Trait Challenge LLD Low-level Descriptor LOI Level of Interest LOSO Leave-One-Subject-Out LP Linear Prediction LPC Linear Predictive Coding LR Likelihood Ratio LSP Line Spectral Pair LSF Line Spectral Frequencies LSTM Long Short-Term Memory LSTM-RNN Long Short-Term Memory Recurrent Neural Network LTAS Long-Term Average Spectrum LTI Linear Time Invariant MCR Mean-Crossing Rate MCT Multi-condition Training MFB Mel-Frequency Band MFCC Mel-Frequency Cepstral Coefficient MIR Music Information Retrieval MLP Multi-Layer Perceptron MPEG Moving Picture Experts Group MRN Mean Range Normalisation xviii Acronyms
- 23. MVN Mean Variance Normalisation NMF Non-negative Matrix Factorisation NN Neural Network PCA Principal Component Analysis PCP Pitch Class Profiles PDA Pitch Detection Algorithm PLP Perceptual Linear Prediction PLP-CC Perceptual Linear Prediction Cepstral Coefficients PMF Probability Mass Function RASTA RelAtive Spectral TrAnsform RASTA-PLP RelAtive Spectral TrAnsform Perceptual Linear Prediction ReLU Rectified Linear Unit RIR Room Impulse Response RMS Root Mean Square RNN Recurrent Neural Network ROC Receiver Operating Characteristic RoP Roll-off Point rProp Resilient Propagation RTF Real-time factor SAD Speech Activity Detection SAL Sensitive Artificial Listener SD Spectral Difference SHS Subharmonic Summation SING Geneva Singing Voice Database SMO Sequential Minimal Optimisation SNR Signal-to-Noise Ratio SPL Sound Pressure Level STFT Short-Time Fourier Transform SUSAS Speech Under Simulated and Actual Stress SVM Support Vector Machine SVR Support Vector Regression TEO Teager energy operator TNR True Negative Rate TPR True Positive Rate UAR Unweighted Average Recall VAD Voice Activity Detection VAM Vera-am-Mittag WAR Weighted Average Recall ZCR Zero-Crossing Rate Acronyms xix
- 24. Symbols General Signal Representation t Time (continuous) n Discrete time index, of a sample, or frame T Period ¿ Discrete time lag (e.g. for ACF) N Number of samples/frames/items/etc. aðtÞ Continuous amplitude of time continuous signal a at time t aðnÞ Continuous amplitude of time discrete signal a at index n ¼ t Ts xðnÞ Discrete amplitude of time discrete signal aðnÞ, or general time series of length N Ts Sampling period in seconds fs Sampling frequency in Hertz fs ¼ 1 Ts Short-Time Processing/Windowing k Discrete frame index ^ n Discrete sample index, relative to a single frame K Total number of frames Nf Frame size in samples Lf Frame size in seconds N ðTÞ f Frame step/period in samples Tf Frame step/frame period in seconds Of Percentage of overlap between two adjacent frames wxxðnÞ Window function of type xx, time domain WxxðmÞ Discrete spectrum of windowing function wxxðnÞ xxi
- 25. Frequency Domain Signals f0 DFT base frequency, DFT frequency resolution f Frequency f ðscaleÞ Frequency expressed in the unit of “scale”, e.g. Hz for “lin” Θscale Frequency scale transformation function from linear f to frequency scale “scale” m Discrete frequency m ¼ f f0 2 ½0; M, bin index m0 Real-valued frequency bin index mðscaleÞ Bin index for spectral scale “scale” FðmÞ Function F which converts the discrete frequency index m to a linear frequency f M Number of discrete frequency bins XðmÞ Discrete (complex) spectrum of signal xðnÞ at discrete frequency m XMðmÞ Spectral magnitude of signal xðnÞ at discrete frequency m XM;normðmÞ (Magnitude) spectral density X`ðmÞ Spectral phase of signal xðnÞ at discrete frequency m XPðmÞ Power spectrum XP;normðmÞ Power spectral density Low-Level Acoustic Descriptors Spectral Bands and Filter Shapes b Spectral band index (for a band spectrum) B Number of spectral bands gðf Þ Spectral filter shape, continuous function in f ΦðmÞ Spectral filter shape, discretised to bins fl Filter bandwidth fc Filter centre frequency for bandpass filters, filter cutoff frequency for high-/low-pass filters fl Lower frequency bound fu Upper frequency bound ml Lower frequency bound, bin index mu Upper frequency bound, bin index fil Weighting factor of lower bound frequency bin fiu Weighting factor of upper bound frequency bin Fundamental Frequency F0 Fundamental frequency (“pitch”) pv Probability of voicing xxii Symbols
- 26. XHðmÞ Subharmonic sum spectrum Nc Number of pitch candidates Energy/Loudness E Signal energy En Normalised signal energy (¼ signal power) Erms Root-mean-square normalised signal energy Elog Logarithmic signal energy El Loudness, auditory-based model El;approx Loudness (narrow-band approximation) ΦfxðnÞg Teager energy operator Spectral Dynamics SDðkÞ Spectral difference between frame k and k 1 SD ðkÞ þ Positive spectral difference between frame k and k 1 Sflux Spectral flux Spectral Statistics Scentroid Spectral centroid Sentropy Spectral entropy Sflatness Spectral flatness Skurtosis Spectral kurtosis SropðnÞ n % spectral roll-off point Sskewness Spectral skewness S Spectral standard deviation Svariance Spectral variance (spectral spread) HM Spectral harmonicity pXðmÞ Spectrum XðmÞ converted to a probability mass function · Hammarberg index ‰fi Alpha ratio PLP/Auditory Spectra XP;audðbÞ Discrete auditory band spectrum EðbÞ Equal loudness weight for spectral band b Symbols xxiii
- 27. Cepstrum/Cepstral Features sðnÞ Vocal tract source signal in the time domain, discrete SðmÞ Vocal tract source signal in the frequency domain, excitation spectrum hðnÞ Impulse response in the time domain, discrete HðmÞ Impulse response in the frequency domain transfer function L In context of MFCC: Liftering coefficient Xfloor Spectral band floor value CðiÞ ith cepstral coefficient Linear Prediction p Linear predictor order ^ xðnÞ Approximation of signal xðnÞ, e.g. through a linear predictor ai Linear predictor coefficients kj Reflection coefficients eðnÞ Error signal (e.g. in linear prediction, residual signal) fij Sum-squares error of linear predictor of order j rðdÞ Autocorrelation coefficient for discrete lag d hinv Impulse response of the inverse vocal tract filter Hinv Impulse response of the inverse vocal tract filter Formants Fi Speech formant frequency, if i 1, e.g. F1 and F2 F ðbwÞ i Speech formant bandwidth for i 1 Harmonic- and Formant-Amplitudes Hi Amplitude of ith spectral harmonic Ai Amplitude of highest spectral harmonic in ith formant range Hij Harmonic amplitude differences/ratios between ith and jth harmonic Voice Quality Jpp Period-to-period jitter (local) Jpp Average (within one frame) period-to-period jitter Jpp Period-to-period shimmer (local) Spp Average (within one frame) period-to-period shimmer xxiv Symbols
- 28. HNRwf Harmonics-to-noise ratio computed via the direct waveform matching method HNRwf ;log Logarithmic harmonics-to-noise ratio computed via the direct wave- form matching method HNRacf Harmonics-to-noise ratio computed via autocorrelation HNRacf ;log Logarithmic harmonics-to-noise ratio computed via autocorrelation Musical s Semitone index S Total number of semitones per octave (for PCP) O Number of octaves Delta/Derived –W i ðnÞ ith order delta regression coefficient at discrete time n; window size W xsmaðnÞ xðnÞ smoothed with a moving average filter (typically 3 frames) Supra-Segmental Features x Feature vector (frame) xðnÞ Time series of feature vectors X Supra-segmental feature vector F Functional Means „ Mean „a Arithmetic mean (amean) „jaj Arithmetic mean of absolute values (absmean) „a þ Arithmetic mean of positive values only (posamean) „a Arithmetic mean of negative values only (negamean) „ ðnzÞ a Arithmetic mean of non-zero values only (nzamean) „g Geometric mean (nzgmean), here: always only of non-zero values „q Quadratic mean (qmean) „rq Root-quadratic mean (rqmean), i. e. square root of „q „f Flatness „x Arithmetic mean of xðnÞ Symbols xxv
- 29. Moments mi ith central (statistical) moment mi ith standardised (statistical) moment Standard deviation 2 Variance Coefficient of Variation Extremes Rx Range of signal xðnÞ xmin Minimum value of signal xðnÞ xmax Maximum value of signal xðnÞ nmin Position n (frame) of minimum value in xðnÞ nmax Position n (frame) of maximum value in xðnÞ dmax;„ Difference between maximum value and arithmetic mean of xðnÞ dmin;„ Difference between minimum value and arithmetic mean of xðnÞ Distributions Pj jth percentile xcentroid Centroid of time domain signal xðnÞ x Absolute amplitude threshold on signal xðnÞ xrel Relative amplitude threshold on signal xðnÞ Regression m Linear regression slope o Linear regression offset a Quadratic regression coefficient 1 (shape, quadratic) b Quadratic regression coefficient 2 (shape, linear) c Quadratic regression coefficient 3 (offset) nv Temporal location (index) of parabola vertex (quadratic regression) ^ xðnvÞ Amplitude of xðnÞ at parabola vertex index nv (quadratic regression) e Normalised (by sequence length) sum-squares (quadratic) regression error ea Normalised (by sequence length) absolute (linear) regression error mleft Left slope of parabola estimated from quadratic regression mright Right slope of parabola estimated from quadratic regression CC Pearson correlation coefficient xxvi Symbols
- 30. Peaks x ðpeaksÞ „ Arithmetic mean of peak amplitudes d ðpeaksÞ „ Mean distance between peaks (temporal) dðpeaksÞ Variance of inter-peak distances x ðminimaÞ „ Arithmetic mean of valley amplitudes d ðminimaÞ „ Mean distance between valleys (temporal) dðminimaÞ Variance of inter-valley distances mj;i Slope (rising) of line connecting a valley i and the following (future) peak j mi;j Slope (falling) of line connecting a peak j and the following (future) valley i nrel Temporal (frame) index relative to length N of segment: nrel ¼ n=N Modelling Neural Networks gðxÞ Squashing function/neuron activation function w Neuron weight vector x Neuron input vector ct LSTM cell state at time t it LSTM cell input gate activation at time t ot LSTM cell output gate activation at time t ft LSTM cell forget gate activation at time t Incremental Processing Npre Number of continuous speech frames required for the detection of the start of a speech segment Npost Number of continuous silence/non-speech frames required for the detection of the end of a speech segment Lmax Maximum allowed speech segment length Lmin Minimum allowed speech segment length N ðiÞ seg Length of the ith speech segment Nwin Window length (in frames) for incremental sub-segmentation Nstep Window step length (in frames) for incremental sub-segmentation Lwin Window length (in frames) for incremental sub-segmentation Lstep Window step length (in frames) for incremental sub-segmentation pw openSMILE data-memory level write pointer pr;i openSMILE data-memory level read pointer of ith reader Symbols xxvii
- 31. Nfree Free space (frames/samples) in openSMILE data-memory level Navail Maxmimum available data items (frames/samples) in openSMILE data-memory level N ðiÞ avail Available data items (frames/samples) in openSMILE data-memory level for ith reader Nxx Same as Nfree and Navail for ring buffer data-memory levels xxviii Symbols
- 32. List of Figures Figure 2.1 Overview of steps of processing (simplified) for general speech and music analysis methods . . . . . . . . . . 10 Figure 2.2 Spectral shapes of two triangular filters designed with a centre frequency of 200 and 600 Mel and a symmetric (on the Mel scale) bandwidth of 200 Mel; Dashed (blue) line with (x) showing Eq. (2.83) and black solid line shows the version from Eq. (2.77) as derived in this thesis based on integration (Eq. (2.73)). . . . . . . . . . . . . 32 Figure 2.3 Mel-band power spectrogram plot of a sample sentence from the AVIC database (Sect. 6.1.3); female speaker, words: “change another color”. . . . . . . . . . . . . . . . . . . . . . 32 Figure 2.4 Cepstrogram plot of a sample sentence from the AVIC database (Sect. 6.1.3); female speaker, words: “change another color” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 Figure 2.5 Auditory spectrogram (based on 26-band Mel-band power spectrum) of a sample sentence from the AVIC database (Sect. 6.1.3); female speaker, words: “change another colour” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 Figure 2.6 RASTA filtered auditory spectrogram (based on 26-band Mel-band power spectrum) of a sample sentence from the AVIC database (Sect. 6.1.3); female speaker, words: “change another colour”. . . . . . . . . . . . . . . 59 Figure 2.7 Left octave (log(2)) scaled spectrum with peak enhancement, smoothing, and auditory weighting applied (X ðoct;wÞ M ); right subharmonic sum spectrum XH. Sample sentence from the AVIC database (Sect. 6.1.3); female speaker, words: “change another colour” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 xxix
- 33. Discovering Diverse Content Through Random Scribd Documents
- 34. the others—as there are villages in the island; and this statement holds equally good whether applied to a large island like St. Christoval or to those of small size as Santa Anna and Ugi. Yet there is not unfrequently to be met a chief who, by the power of his wealth or by the number of his fighting men, assumes a degree of suzerainty over the less powerful chiefs in his vicinity. Thus, the influence of Gorai, the Shortland chief, is not only dominant over the islands of Bougainville Straits, but extends to the adjacent coast of the large islands of Bougainville and Choiseul, and reaches even to Bouka, more than a hundred miles away. The small island of Simbo or Eddystone, the Narovo of the natives, is under the sway of a powerful chief who resides, together with nearly all his fighting men, on an islet bordering its south-east side. His influence extends to the neighbouring larger islands, and is probably as despotic as that of any of the numerous chiefs with whom I was brought into contact. I might mention other instances in this group where a comparatively small island becomes the political centre of a large district. Similar instances are familiar amongst the other Pacific archipelagos, and notably in the case of Bau in Fiji; and they may all be attributed to the fact that the coast-tribes are of more robust physique and of more enterprising character than the inhabitants of the interior of the larger island, or “bush men” as they are often termed. The large island of St. Christoval is divided amongst numerous tribes between which there are constant feuds, each tribe having its own chief. A wide distinction exists between the inhabitants of the interior and those of the coast; and an unceasing hostility prevails between the one and the other. The distinction often extends to language, a circumstance which points to a long continuation of these feuds; and from it we may infer that the isolation has continued during a considerable period. The bush-tribes find their best protection on the summits of the high hills and on the crests of the mountain-ridges which traverse the interior of the island. I passed one night in the bush-village of Lawa, which is situated on a hill-top about 1,400 feet above the sea near the north coast of St. Christoval. As I was in a locality where probably no white man had been before, the novelty
- 35. of my situation kept me awake the greater part of the night; and very early the next morning I rose up from my mat in the tambu- house to view, undisturbed, the interior region of the island. It was a gloomy morning. Thin lines of mist were still encircling the loftier summits or lingering in the valleys below. Here and there on the crest of some distant hill a cluster of cocoa-nut palms marked the home of a bush-tribe effectually isolated by deep intervening valleys from the neighbouring tribes. I gazed upon a region which had for ages worn the same aspect, inhabited by the same savage races, the signs of whose existence played such an insignificant part in the panorama laid out before me. Standing alone on this hill-top, I reflected on the deeds of barbarity which these silent mountains must have witnessed “in the days of other years,” deeds which are only too frequent in our own day when the hand of every tribe is against its neighbour, and when the butchery of some unsuspecting hamlet too often supplies the captors with the materials for the cannibal feast. By the unusual success of their treachery and cunning—the two weapons most essential to savage warfare in St. Christoval as well as in the other islands—some chiefs have acquired a predominance over the neighbouring villages, and their name inspires terror throughout the island. Amongst them, I may mention Taki, the chief of the large village of Wano on the north coast of this island. He has obtained the double reputation of being a friend to the white man and of being the most accomplished head-hunter in St. Christoval; and, as may be readily imagined, the efforts of the Melanesian Mission, by whom a station has been for many years established in this village,[5] have been greatly retarded by the indifference of this powerful chief. The resident teacher in the village was his own son, who had been selected by Bishop Selwyn and had undergone the usual training of teachers in Norfolk Island. I regret to write that he greatly lapsed during our stay in the group, that he appears to have accompanied his father on a head-hunting foray, and that he finally met with an untimely fate, being so severely wounded by a shark when fishing on the reef that he died a few hours afterwards. Taki,
- 36. although not a Christian convert, was fond of displaying his connection with the Mission. He showed me a certificate which he received from Bishop Patteson in July, 1866; and in fact he is always ready to do the honours of his village to the white man. Of his head- hunting propensities, Captain Macdonald, an American trader resident in Santa Anna, told us the following tale: Not long before the arrival of H.M.S. “Lark” in the Solomon Islands, he was sailing along the St. Christoval coast, when he met Taki in his war-canoe proceeding on one of these expeditions. He endeavoured to place hindrances in the chief’s way by telling him that he had native- traders living at the different places on the coast where he intended to land. But it was to no purpose. Taki saw the ruse, and taking it in good part remarked to Captain Macdonald that he had apparently a large number of natives trading for him. Waiting patiently until some unfortunate bushmen ventured down on the reefs to fish, the Wano chief surprised them, slaughtered many and carried the living and the dead in triumph to his village. When Mr. Brenchley visited this village in H.M.S. “Curacoa” in 1865, he saw evidence of a head- hunting foray, in which probably Taki had taken part in his youthful days. The skulls of 25 bushmen were observed hanging up under the roof of the tambu-house, all showing the marks of the tomahawk.[6] In our time, this chief conducted his forays less openly, and I saw no evidence of his work in the tambu-houses of his village. [5] The Rev. J. Atkin was resident at Wano in 1871, shortly before he met his death with Bishop Patteson in Santa Cruz. [6] “Cruise of H.M.S. ‘Curacoa’” (p. 267); by J. L. Brenchley, M.A. The practice of head-hunting, above referred to, prevails over a large extent of the Solomon Group. The chiefs of New Georgia or Rubiana extend their raids to Isabel, Florida, and Guadalcanar; and thus perform voyages over a hundred miles in length. Within the radius of these raids no native can be said to enjoy the security of his own existence for a single day. In the villages of Rubiana may be seen heaps of skulls testifying to the success of previous
- 37. expeditions. Captain Cheyne, when visiting Simbo or Eddystone Island in 1844, found that the natives had just returned from a successful expedition, bringing with them 93 heads of men, women, and children. In these expeditions, he says, they sometimes reached as far as Murray Island which lies about 135 miles to the eastward. [7] Their reputation, however, had extended yet further, since D’Urville, who visited Thousand Ships Bay in 1838, tells us that the Isabel natives knew the land of Simbo and pointed to the west to indicate its direction.[8] The Rev. Dr. Codrington, in referring to these head-hunting raids,[9] remarks that the people of the south-west part of Isabel have suffered very much from attacks made on them year after year by the inhabitants of the further coast of the same island and of neighbouring islands, the object of these attacks being to obtain heads, either for the honour of a dead or living chief or for the inauguration of new canoes. He observes that a new war canoe is not invested with due mana, i.e., supernatural power, until some man has been killed by those on board her; and any unfortunate voyagers are hunted down for the purpose on the first trip or afterwards. The Rubiana natives are said to have introduced head- hunting and human sacrifices into the neighbouring islands. They carry off not only heads but living prisoners, whom they are believed to keep, till on the death of a chief, or launching of a canoe, or some great sacrifice, their lives are taken. [7] “A Description of Islands in the Western Pacific Ocean” (p. 66), by Andrew Cheyne, London, 1852. [8] “Voyage au Pole Sud,” Paris, 1843; tom. v., p. 31. [9] Journal of Anthropological Institute, vol. x., p. 261. White men have sometimes been the victims of these head-hunting expeditions. As is well known, Lieutenant Bower, of H.M.S. “Sandfly,” met his death, together with the greater number of his boat’s crew, on the islet of Mandoleana, in 1880, at the hands of a similar expedition undertaken by the Florida natives. Kalikona, the most influential chief of the Florida Islands, was freed from implication in this tragedy mainly through the efforts of Bishop Selwyn, to whose
- 38. influence the subsequent surrender of the five natives concerned in the raid was chiefly due. More often than not, these head-hunting forays are unconnected with cannibalism, the mere possession of skulls being the principal object of the expedition. In some islands, there is a rude idea of justice perceptible in this practice. It is the custom in the eastern islands of the group to place out head-money for the head of any man who may have rendered himself obnoxious to any particular village. The money—a considerable amount of native shell-money—may be offered by the friends of a murdered man for the head of the murderer. Months, sometimes years, may elapse before the deed is accomplished and the money paid. The task is generally undertaken by a professional head-hunter, such as we met in the person of Mai, the second chief of the village of Sapuna, in the island of Santa Anna. To make a thorough examination of the home and surroundings of his victim, and to insinuate himself into that intimacy which friendship alone can give him, are necessary initiatory steps which only the cunning head- hunter can know how to carry to a successful issue. Time is of no moment. The means employed are slow, but the end is none the less secure; and when the opportunity arrives, it is the friend of months, if not of years, who gives the fatal blow. In the above description of the head-hunter, I have had before my mind some of the reminiscences of Captain Macdonald, to whom I have before alluded. By his judicious treatment of the natives in the eastern islands, he has acquired a powerful influence for good amongst them; and it is to his past discretion that many a white man, myself among the number, has owed his safety when landing on St. Christoval. When this island was being surveyed by the officers of H.M.S. “Lark,” in 1882, we learned that there was head-money out for a white man’s head in a district on the north side and nearly opposite Ugi. It appeared that about a year before a fatal accident had occurred on board a trading-vessel through a revolver going off unexpectedly and killing a native belonging to the district. It was the current opinion of
- 39. resident traders that sooner or later the required head would be obtained. As characteristic of a trader’s experience in these islands, I may add that on one occasion when visiting Mr. Bateman, a trader residing then on the north coast of Ugi, I was told by him that about a month before a friendly Malaita chief had arrived in a large canoe at Ugi with the information that head-money had been offered by another Malaita chief for the head of a white man. The chief who brought the news advised Mr. Bateman to remove his residence to the interior of the island; and the natives in his vicinity were very solicitous that the warning should be heeded. I learned from Mr. Stephens, who has resided on Ugi for several years, that on one occasion when he was resident on Guadalcanar, on returning from an excursion up the bed of one of the streams, a message was received from the chief of a village in the interior warning him not to make any more similar excursions or he would take his life. The chief of the village, under whose protection Mr. Stephens was residing, took up the matter as an insult to himself; and sent a reply to the effect that if the neighbouring chief wished to remain on terms of amity with him, he should at once send a head in atonement for the threats directed against the white man. A day or two afterwards, Mr. Stephens saw the head, which had been duly sent. The little island of Santa Anna, although but 21⁄2 miles in length, supports two principal villages, Otagara and Sapuna, which are as often as not at war with each other, although only separated by the breadth of the island. Such was the state of affairs during one of our visits to Port Mary in this island; and the fact that the natives of the two villages were connected by inter-marriages did not act as a deterrent in the matter. Through the restless spirit of Mai, the head- hunter before referred to, some old grievance had been dug up, the murder, I believe, some years before of the brother of Mai by the Otagara natives. The outcome of it was that in the middle of the night all the fighting men of Sapuna assembled at the tambu-house of Mai, and started off along the coast to pounce upon their fellow
- 40. islanders on the other side. The utmost that could have happened would have been the slaughter of some unsuspecting man or woman on the skirts of the village: but, as it chanced, a thunderstorm with heavy rain overtook the party when near their destination; and this dampened their courage to such a degree that they returned to their own village with the excuse that the rain, by running down their faces, would have hindered them in throwing their spears and avoiding those of their opponents. On the following day, Mai led a party of Sapuna men to make another attack, and on returning in the afternoon from one of my excursions into the interior of the island, I learned that the party had returned triumphant, having killed one of their neighbour’s large pigs, an act which is regarded as a “casus belli” in native politics. In the person of Mai, we have a typical example of a Solomon Island head-hunter. The cunning and ferocity which marked his dealings, were sufficiently indicated in his countenance and his mien. He had established for himself the position of war-chief in his village of Sapuna, the reigning chief being of a more peaceable disposition. During one of our visits to this island we found that this war-chief had been very recently displaying his heroism in the most approved native fashion. He had led a war-party across to Fanarite on the opposite coast of St. Christoval, to avenge the death of a fugitive from a labour vessel who, having escaped at Santa Anna, subsequently found his way to Fanarite where he was killed. The excuse, although somewhat circuitous, was quite sufficient for Mai, who in his disinterestedness thought more of this chance of gaining new laurels than of the untimely end of the native whose death he was so eager to avenge. Having reached the part of the coast where this man had been killed, the war-party lay in ambush and slaughtered a chief and two women as they were returning from their yam patches; whilst they severely wounded another woman who escaped into the bush with a spear through her back. Having dipped their weapons in the gore of their victims, Mai and his party returned to Santa Anna. I was sorry to learn that a native, named Pukka-pukka who had served in the “Lark” as an interpreter during
- 41. the previous year, had taken an active part in this expedition. It appeared that the chief had aimed at him, but his musket missed fire, when Pukka-pukka shot him through the back with his snider. The scene of the tragedy was familiar to me, as I had landed there the year before. Pukka-pukka, who is a sensible young man and of by no means a bloodthirsty disposition, did not like my taking him to task for the part he took in this raid; and he protested more than once in a somewhat injured tone that his people did not fight without good cause. In his case, I felt confident that he was not tempted by the mere love of bloodshedding, the truth being that through the able tutorship of Mai, all old feuds are kept alive in the minds of the young men of the village, who, in their desire to distinguish themselves, come to regard such grievances as fair grounds for war. We soon learned that the Fanarite natives would seize the first opportunity to retaliate; and that head-money to a large amount had been offered for the head of a native of Santa Anna, and particularly for the head of Pukka-pukka. The chiefs of the islands of Bougainville Straits possess far greater power over their peoples than that which is wielded by most of the chiefs we encountered at the St. Christoval end of the group. At Santa Anna and at Ugi, the position of the chief is almost an empty honour; and some man of spirit, though not of principle, such as Mai in the former island and Rora at Ugi, usurps by his fighting prowess a large share of the power. On the St. Christoval coast I met several such chiefs, who possess no influence beyond their own district, and often very little in that. Occasionally, as I have before observed, a chief is found who, like Taki at Wano, exercises a powerful influence over the less pretentious chiefs of neighbouring islands and districts. Some of the Guadalcanar chiefs are very powerful; but with them I had no personal intercourse; and I prefer to confine my remarks to those portions of the group with which I became acquainted. Returning, then, to the chiefs of the islands of Bougainville Straits, I may enumerate them in their order of importance—Gorai in the Shortland Islands, Mule at Treasury, Kurra-kurra and Tomimas in Faro or Fauro, and Krepas at Choiseul Bay. There is constant
- 42. communication between the natives of these islands, more particularly between those of Treasury, the Shortlands, and Faro, the distances between the islands varying between 15 and 25 miles. Intermarriages are frequent between the natives of these islands. They all speak the same language; and not uncommonly a man shifts his home from one island to another. The chiefs are all connected either by blood-relationship or by marriage, and together form as powerful an alliance as might be found in the whole group. Visits of condolence are exchanged in times of bereavement between the chiefs; and presents are conveyed from one to another. On one occasion we carried a present of sago from Mule to Gorai; and I have on more than one occasion during our passages between these islands been made the bearer of a message from chief to chief. 1 2
- 43. 1. Gorai, his principal Wife, and his Son Ferguson. 2. Four of the Wives of Mule. [To face page 21. Gorai, the well-known Alu chief, Alu being the name of his principal island, exercises a kind of suzerainty over the neighbouring chiefs. But his reputation and influence extend far beyond the islands directly or indirectly under his rule. From Treasury northward and eastward, throughout the Shortlands, across the straits to Choiseul Bay, through Faro, and along the coast of Bougainville, extending even to Bouka, his influence is predominant. Masters of vessels, recruiting labour on the coast of Bougainville, have a sufficient guarantee for the good behaviour of the natives of the places they visit, if they have been fortunate enough to secure the presence on board of one of the sons of Gorai. This chief has been the trusted friend of the white man for many years. On our first visit to Alu we were therefore prepared to think favourably of him. We found him on the beach, surrounded by a considerable number of his people. Shaking hands with us, he told us in his imperfect English that he was a friend of the white man. Rather beyond middle age, and somewhat shorter than the average native, he has an honest, good- humoured expression of countenance, which at once prepossessed us in his favour. Whilst seated in the dingy interior of one of his
- 44. houses, surrounded by several of his wives, Gorai related to us the story—well known to all acquainted with the Solomon Group—of his reprisal a few years before on the natives of Nouma-nouma, a village on the east coast of Bougainville, for the murder of Captain Ferguson of the trading steamer “Ripple.” The master of the “Ripple” was an old friend of Gorai, and traded extensively with him. On hearing the news, the chief mustered his men and despatched them in canoes, under the command of his eldest son, to the scene of the massacre, about a hundred miles away. The natives of the offending village were surprised, and about twenty of them were killed, including men, women, and children—“all same man-of-war,” as Gorai too truthfully observed. One of the chief’s sons has received the name of the unfortunate master of the “Ripple;” and I may here refer to the good name which Captain Ferguson has left behind him, not only amongst the natives of the Solomon Islands, but also amongst his fellow-traders in those seas. The inhabitants of the Shortland Islands, Gorai’s immediate rule, live in great awe of their chief; and the number of natives who gathered round us when we first met the chief showed us by their manner that in the friendship of the chief the white man possessed the goodwill of his subjects. We were unable to see very much of the mode of exercising his power; but I suspect that Gorai, like other chiefs, places but little value on the lives of his people. Punishment is summarily dealt by the spear or the tomahawk; and I learned from natives of the adjoining islands that the offence may be of a very trivial nature. On one occasion, Gorai took me in his war-canoe on a geological excursion to the north-west side of Alu. During our return, the sun set when we were about twelve miles from the ship, and left us to pursue our way in the darkness. Seated alongside the chief on the second bow thwart of the canoe, I could not help reflecting how many times he must have occupied the same seat in his war-canoes when engaged in those expeditions which have made his influence dominant on this part of the group. On our way we skirted the beach of an islet on which were squatting a party of Alu natives who had gone there to fish. Although we passed a few yards from these men,
- 45. not a word of recognition was exchanged. The sight of a large war- canoe with Gorai and a white man in the bow passing them in the dusk of evening must have been a novel one to them, yet neither they nor our men exchanged a word. There they sat squatting motionless on the beach, and we passed them in silence. Gorai subsequently explained to me that the reason of this was that the men were “too much fright,” or rather awed, by the presence of their chief. The chief of the Shortland Islands has two or more elderly men who act as his ministers. Many years ago he was living at Treasury, of which island he was chief; but being unwilling to take part in the hostility displayed by the Treasury natives towards the white men, he left the island under the chieftainship of Mule, the present chief, who still remained in some degree under the rule of Gorai. The Alu chief takes a pleasure in asserting that he is “all same white man,” at the same time deprecating the inferior position of his race with the remark, “White man, he savez too much. Poor black man! He no savez nothing.” I now come to Mule, the Treasury chief, who numbers amongst his wives a sister of Gorai, Bita by name; whilst the Alu chief has returned the compliment by making Mule’s sister, Kai-ka, the principal amongst his hundred wives. Mule, also known as Mule- kopa, has rather the appearance and build of a chief of one of the more eastern Pacific groups. He has a sedate expression of countenance, a prominent chin, and strongly marked coarse features. A large bushy head of hair adds to the dignity of his appearance; and his powerful limbs, depth of chest, breadth of shoulders, and greater height distinguish him pre-eminently from his people. His rule is as despotic in Treasury as that of Gorai in the Shortlands; and he maintains his sway rather by the fear he inspires than by possessing any feeling of respect on the part of his subjects. On more than one occasion I have heard the natives use threatening language towards their chief, when he had made some arbitrary exercise of his power. He had a habit of sending away to the bush
- 46. any native who from his superior knowledge of English seemed to be supplanting him in the intercourse with the ships that visited the harbour. Even his right-hand man, who prided himself on his name of Billy, experienced his wrath on one occasion in this manner. Like other chiefs, Mule is grasping and covetous, shortcomings which are rather those of the race than of the individual. Although of the chiefs of Bougainville Straits I liked him the least, the contrast was rather due to the exceptionally good estimate we had formed of his fellow chiefs. The visits of H.M.S. “Lark” to this island have been the means of removing the very bad reputation which the natives had deservedly possessed: and I would especially invite the attention of my readers to the history of this change in the attitude of these natives towards the white man. Captain C. H. Simpson, who visited this island in H.M.S. “Blanche” in 1872, described its people in his report to the Admiralty,[10] as being “the most treacherous and blood-thirsty of any known savages;” and the officers employed in making a sketch of the harbour had ample evidence of their ferocity. About seven years before, the natives had cut out a barque and had murdered her crew of 33 men. Previously they had captured several boats of whalers visiting the islands, and had massacred the crews. The Treasury natives were always very reticent to us when we tried to learn something more of the fate of the barque; but we learned little except that she was American, and was named “Superior.” The captain, whose name the natives pronounced “Hoody,” was carried away into the interior of the island and killed, and the scene of his murder was once pointed out to Lieutenant Oldham when crossing the island. As Captain Simpson charges the natives with cannibalism, there can be little doubt of the ultimate fate of the crew of the American barque. In the interval between the occurrence of this event and the arrival of the “Blanche,” no vessel had anchored in the harbour, the ships always heaving-to off the north coast, where the natives resided when Captain Simpson visited the island. Treasury retained its bad reputation up to the date of our visit; and but few traders had much knowledge of the place, as they generally gave the island a wide
- 47. berth. We met but one man who spoke well of these natives, and he was Captain Walsch of the trading schooner “Venture.” All others gave them the worst of characters: and led me to believe that my acquaintance with Treasury would not extend beyond the deck of H.M.S. “Lark.” When Lieutenant Oldham first visited this island in May, 1882, he had every reason to place but little confidence in the natives; and in truth we all thought that the appearance and behaviour of the natives justified the treacherous reputation which they had obtained. Only two days were spent there, but no landing was effected: the chief made no response to the invitations to visit the ship; and we left the harbour without much feeling of regret. In June of the following year we again visited this island; and if the same procedure had been followed we should have been a very long time in gaining the confidence of the natives. Lieutenant Oldham, however, paid an official visit to the chief, accompanied by Lieutenant Malan and myself. Mule and one of his sons returned the visit within a couple of hours. Presents were exchanged; and the foundation of mutual confidence was thus laid. The result may be briefly stated. In a few days I was rambling all over the island, usually accompanied by a lively gathering of men and boys. An intimacy was established with the natives, which lasted until we bade farewell to the group in the following year; and the return of the “Lark” from her cruises was always a cause of rejoicing amongst the natives. The men of the ship were known by name to most of the people of the island: whilst Mr. Isabell, our leading-stoker, made a deep impression upon them by his readiness to employ his mechanical skill for their various wants, so much so that Mule offered, if he would remain, to make him a chief with the usual perquisite as to the number of his wives. For my own part, I reaped the full benefit of our amicable relations with the natives; and for the proof of this statement I must refer the reader to the remarks on my intercourse with them, and to my observations on the geology, botany, and other characteristics of the island. [10] “Hydrographic Notices, Pacific Ocean,” 1856 to 1873 (p. 106).
- 48. Coming now to the chiefs of Faro or Fauro Island, I must mention more particularly Kurra-kurra the chief of Toma, and Tomimas the chief of Sinasoro, Toma and Sinasoro being the two principal villages of the island. Kurra-kurra is, I believe, a half-brother of Gorai. He has not, however, the same dignity of manner, and has resigned most of his power into the hands of his son Gorishwa, a fine strapping young man. Both father and son are friends of the white man. Tomimas, the Sinasoro chief, also related to Gorai, is somewhat taciturn even with his own people, but a chief to be thoroughly trusted. On one occasion whilst assisting Lieutenant Heming and myself in demolishing our dinners in a tambu-house at his village, Tomimas broke a long silence by informing us through a native interpreter that the men of Sinasoro were very good people, that they did not kill white men, and that their chief was like Gorai. It is needless to write that we appreciated the good intention, though hardly the elegance of the chief’s solitary remark. In the following year, when I was returning from a botanical excursion to the peak of Faro, I received an invitation from Tomimas to visit him on the side of the harbour opposite to the village. The chief, who awaited me on the beach, received me cordially, telling me through one of the natives, who could speak a little English, that he had collected for me the fruits and leaves of the “anumi”—a tree of the genus Cerbera—which he had heard I had been anxious to find. The kindly manner of the old chief attracted me towards him, and I sat down, as he wished me, by his side on the log of a tree, having first presented him with a large knife which greatly pleased him. Close by, stood his four wives, to whom he introduced me, pointing out to me the mother of his eldest son Kopana, an intelligent young man of about twenty- two. A bunch of ripe bananas was laid beside me, of which I was bidden to partake. This was followed in a short time by a savoury vegetable broth, which the chief brought with his own hands in a cooking-pot. It was especially prepared for me on their learning that I had found the plant (an aroid, Schizmatoglottis) in my excursions. There was the spirit of true politeness displayed in the manner of the chief and his wives, as they endeavoured to show that in the exercise of their simple hospitality they were receiving, instead of
- 49. conferring, an honour. I felt that I was in the presence of good breeding, although sitting attired in a dirty flannel suit in the midst of a number of almost naked savages. My own party of Sinasoro natives, who had been fasting for many hours, politely asked me to partake of their meal which the generosity of the chief had prepared, before they thought of touching it themselves. I of course complied with their request by tasting a cooked banana, when, this piece of etiquette having been duly observed, they attacked the victuals without ceremony. Such was my pleasing experience of this Faro chief. During the survey of this island, the natives showed every disposition to be friendly towards us. In my numerous excursions I always met with civility, and frequently with unexpected acts of kindness; and I soon became known to them by the name given to me by the Treasury natives, “Rōkus” or “Dōkus.” The principal chief of the district, immediately north of Choiseul Bay, is named “Krepas.” Several years before he had been living at Faro, which he left on account of the death of all his wives. When we first visited Choiseul Bay in September, 1883, we found the natives very coy in approaching us, on account of the reprisal of H.M.S. “Emerald,” two years before, on the people of the neighbouring village of Kangopassa for the cutting out of the trading-vessel “Zephyr,” and the murder of a portion of her crew. After two days, however, Lieutenant Oldham succeeded in removing their suspicions, and the chief came on board. Subsequently Krepas and his son, Kiliusi, accompanied me in a canoe during my ascent of one of the rivers that empty themselves into this bay. I found the chief and his son very useful guides, and was prepossessed in their favour. On our return to Treasury, I was surprised to learn from Billy, Mule’s prime minister, as we termed him, that Krepas was a practised cannibal, and would not think much of killing a white man. Billy was deeply impressed by the circumstance of my having shared my lunch with the chief of Choiseul Bay, about two miles up one of the rivers. It was in this bay that the French navigator, Bougainville, intended to
- 50. anchor his ships in 1768, being opposed by the hostility of the natives. The boats, which had been sent in to find an anchorage, were attacked by 150 men in ten canoes, who were only routed after the second discharge of fire-arms. Two canoes were captured, in one of which was found the jaw of a man half-broiled. The number of shoals, and the irregularity of the currents prevented the ships coming up to the anchorage before night fell; and Bougainville, abandoning his design, continued his course through the Straits.[11] The description which the French navigator gave of these natives in 1768, applies equally well to those of the present day. When H.M.S. “Lark” revisited Choiseul Bay in October, 1884, not a single native was seen; so that it would behove future visitors to be very cautious in their dealings with these natives. Whilst off the coast north of this bay, a fishing-party of half-a-dozen men came off to the ship from the village of Kandelai; but they showed great suspicion of us. They would not come alongside for some time; and when a present of calico was flung to them at the end of a line, they were divided amongst themselves whether to come and take it, some paddling one way and some another. At length they took the present and came alongside, but did not stay long, and soon paddled towards the shore, their suspicions by no means allayed. What had happened to cause this change of attitude, we could not learn. Evidently, the good impression which we had left behind us a year before, had borne no fruit. Probably, some inconsiderate action on the part of the crew of a trading-vessel had undone our work. [11] “Voyage autour du Monde,” 2nd edit. augm. vol. II., Paris, 1772. The professional head-hunter of the eastern islands of the group does not appear to be represented amongst the islands of Bougainville Straits. Raids are occasionally made on the villages of the adjoining Bougainville coast, but more, I believe, for the purpose of procuring slaves, than from the mere desire of fighting. There is, however, frequent friendly communication between the natives of the islands of the Straits and those of certain Bougainville villages, the former usually exchanging articles of trade for spears and
- 51. tortoise-shell, and acting as middle-men in the traffic with the white men. It is however singular that the natives of the Straits trade with different villages on the Bougainville coast; and that, although on usually such friendly terms with each other, they are often on terms of hostility with the particular Bougainville village with which their neighbours trade. Thus, Mule, the Treasury chief, trades with the people of the village of Suwai, over which his brother Kopana is chief. Gorai, the Alu chief, on the other hand, is at war with the natives of Suwai, but maintains friendly communication with Daku, the chief of the village of Takura, and with Magasa the chief of the harbour of Tonali. Whilst spending a night at Sinasoro with Lieutenant Heming and his party, I with the rest had to share the tambu-house with a party of ten natives from Takura. They had come across for pigs and taro. The natives of the adjoining coast of Bougainville, possessing a different language, are not able to make themselves understood by the people of the Straits except by interpreters. I have seen one of these natives just as little able to make himself understood by the natives of Faro, as if he had been suddenly removed to some very distant country instead of only 30 miles away. I have previously referred to the close friendship which usually prevails between the inhabitants of the islands of Bougainville Straits, linked together as they are by inter-marriages and by the possession of a common language. But in the calmest seas there are occasional storms; and I will proceed to relate an extraordinary chain of events which came more or less under our observation whilst in this portion of the group. Shortly before our return to Treasury in April, 1884, there had been a terrible domestic tragedy, which at one time threatened to embroil all the chiefs of the Straits in actual war. It appeared that Kopana, the eldest son of Gorai had, in a fit of temporary madness, shot one of his wives dead with his rifle, the unfortunate woman being a daughter of Mule, the Treasury chief. On hearing the news, Mule at once crossed over to Alu to exact vengeance on Kopana; but Gorai would not permit him to harm his son; and it was arranged between the two chiefs that Mule
- 52. should be allowed to shoot one of the other wives of Kopana, as the price of blood. Early one morning the Treasury chief, armed with his snider rifle, took his way in a canoe up a passage I had often traversed in my Rob Roy, and surprising his selected victim at work in a taro patch, he shot her dead. At the same time he wounded her male attendant, an elderly native named Malakolo, the bullet passing through the left shoulder-joint from behind. When I saw this man six or seven weeks afterwards, he was fast recovering from the injury, although with a useless limb. Kopana, who is a headstrong son and beyond his father’s control, naturally resented this act of Mule, and appears to have meditated a descent on Treasury. Collecting his followers and the remainder of his wives, he disappeared on what was given out as a tortoise-shell expedition. We found the Treasury people in a great dread of the daily arrival of Kopana; and I had some difficulty in getting natives to accompany me in my excursions about the island. They did not care to leave the vicinity of the village; and I found many of the bush-paths familiar to me in the previous year partly overgrown. Apparently through a sense of shame, Mule and his natives avoided telling us anything about the act of retaliation; they were, however, loud in their endeavours to cast aspersions on Kopana. On our arrival at Alu, we learned the truth from Gorai to whom Mule had sent a native, who took a passage with us, asking him not to be too communicative in case we made inquiries. As it happened, however, the Treasury native was kept on board, and Lieutenant Oldham, on landing, learned the part Mule had played. Kopana was apparently quite conscious of his own responsibility in the matter, as he had left a present with Gorai to be given to the captain of any man-of-war who should come to punish him. Thus closed the first scene of this tragedy. Whilst we lay at anchor off Gorai’s village, it was evident that there was trouble brewing. The natives accompanying me in my geological excursions carried arms contrary to their usual practice. On the same day the two principal villages were found deserted; and Gorai shifted his residence to another islet. Rumours became rife that the Treasury and Shortland natives had met with bloodshed; but the
- 53. men we questioned made so many wilful misstatements that it was impossible to learn what had really happened. At length the truth came out. Being in Gorai’s house one morning, I was told by the chief that his son had been attacked five days before by the Treasury natives on the islet of Tuluba, off the west coast of Alu, that Kopana’s canoe had returned without his master, bringing a man and a woman badly wounded, and that he shortly expected the return of two large war-canoes which he had sent to the scene of the encounter. These two canoes returned whilst I was talking to the chief on the beach, bringing a few more survivors but without Kopana. The old chief then took it for granted that his eldest son was dead, and in telling me so showed no emotion whatever. In the evening, however, we learned, to our astonishment, that Kopana had returned, having not been engaged in the fray. It seemed that at the time of the encounter he was on a neighbouring islet. After some difficulty, I was able to get an account of the affair. Two Treasury war-canoes, it appears, attempted to land at Tuluba Islet one evening, where the crews were going to encamp for the night. Ostensibly the Treasury men were on their way to Bougainville to buy spears; but since they were led by Olega, the brother of Mule and the fighting-chief of the island, it is probable that they were intending a descent on Alu from this islet of Tuluba. When the Treasury men discovered Kopana’s party were already there, the fighting at once began. During the conflict, for which the Alu natives were ill-prepared, seeing that they were largely composed of Kopana’s wives, one of the Treasury canoes was dashed to pieces on a reef and all the occupants were thrown into the water. In this unequal contest, the Alu natives had a man and a woman killed and a man and a woman wounded, both the women being wives of Kopana. In addition four other of Kopana’s wives were captured by the Treasury men, who returned to their own island in the remaining canoe with a loss of four men wounded, of whom one subsequently died.
- 54. The unfortunate wives of Kopana had indeed borne the brunt from the very beginning. Within two months, three of them had suffered violent death, one of them was wounded apparently beyond recovery, and four had been carried off prisoners to Treasury. The singular feature of this breach between the Treasury and Alu natives, was that the animosity of the former was directed against Gorai’s eldest son and not against the old chief, his father, who did not think it incumbent on him to interfere except for the purpose of pacifying the two parties. I visited the two wounded brought back to Alu. Five days had already elapsed since the fight, and I found the wounds of both in a horrible condition. The wife of Kopana had a severe tomahawk wound of the thigh just above the knee, smashing the bone and implicating the joint. The man had a rifle-bullet wound through the fleshy part of the thigh and a pistol-bullet wound in the opposite groin. Nothing had been done in either case, and after the lapse of five days in a tropical climate, the condition of the wounds could be scarcely described. I was allowed to do but little, and considered recovery in either case most improbable. Both, however, recovered to my great astonishment. I found afterwards, on visiting the wounded at Treasury, that one man had been shot through the elbow-joint by one of his own party. The subsequent events in connection with this outbreak of hostilities in the Straits may be soon related. Although there was now open war between Alu and Treasury, it assumed a passive character, each side awaiting or expecting an attack from the other. Gorai was much concerned at this turn of events, seeing that, as he told me, he thought he had come to an amicable arrangement with Mule when he allowed him to take the life of one of his son’s wives. The canoe- houses at Alu were usually filled during the day by a number of natives, all carrying their tomahawks and debating on the topic of the day. In the midst of them I once found Gorai talking in his quiet way to an attentive circle of armed natives. In the meanwhile the Treasury natives held a feast in celebration of their success; and the
- 55. four wives of Kopana were distributed about the village, but they experienced no ill treatment. In a few weeks the animosity displayed between the peoples of the two islands began to cool down; and it soon became evident that the war was one only in name. At length peace was once more restored. In the beginning of October a number of Treasury natives came over to the west coast of Alu where Gorai was then residing, bringing with them Mule’s principal wife, Bita, the sister of the Alu chief, together with a large present of bananas, taro, and other vegetables; and lastly, what was the most significant act of all, they brought with them the four wives of Kopana who had been captured on the islet of Tuluba. Gorai told me that amity was now perfectly restored, and that he was going to exchange visits with the Treasury chief to confirm the compact. Fortunately for the happiness of the natives of Bougainville Straits, war rarely disturbs the peaceful atmosphere in which they live. I cannot doubt that, in the lives of the natives of these straits, we have the brighter side of the existence of the Solomon Islander; and this result may, I think, be attributed in the main to the influence of Gorai, the Alu chief, who in his intercourse with white men, not always the best fitted to represent their colour, as I need scarcely remark, has learned some lessons in his own crude way which he could hardly have learned under any other conditions. Natives of the islands of the Straits can count with some confidence on the tenure of their lives, but this is simply due to the influence of the name of the Alu chief. And yet, however secure the surroundings of a native may be, he will never be entirely off his guard. Suspicion is a quality inherent in his mind, and it shows itself in most of the actions of his life. Even of those natives, who, in the capacity of interpreters, lived on board the ship for weeks together, one was always keeping watch over his comrades during the long hours of the night whenever we were at any anchorage away from their own island; and I have been told by the officers in charge of the detached surveying parties, that even after a hard day’s work in the boat, they have found their natives keeping a self-imposed watch during the night.
- 56. I pass on now to the subject of the power of the “tambu,” or “taboo” as it is more usually termed. The tambu ban constitutes the real authority of a petty chief in times of peace. In the eastern islands, the tambu sign is often two sticks crossed and placed in the ground. In such a manner, the St. Christoval native secures his patch of ground from intrusion. In the islands of Bougainville Straits, posts six to eight feet in height, rudely carved in the form of the head and face, are erected facing sea-ward on the beach of a village to keep off enemies and sickness. Similar posts are erected on the skirts of a plantation of cocoa-nut palms to warn off intruders. On one occasion, whilst ascending the higher part of a stream in Treasury, my natives unexpectedly came upon the faint footprint of a bushman; and my sheath-knife was at once borrowed by the chief’s eldest son, who happened to be one of the party, to cut out a face in the soft rock as a tambu mark for the bushman, or in other words to preserve the stream. I have only touched on the exercise of the right of tambu in its narrowest sense. Scattered about in the pages of this work will be found numerous allusions to customs which would be comprised under this head in its widest meaning: for the power of the tambu is but the power of a code which usually prohibits and rarely commands; and in enumerating its restrictions and defining its limits, one would be in reality describing a negative system of public and private etiquette. It is worthy of note, that the term “tambu” is not included in the vocabulary of the language of the natives of Bougainville Straits, its equivalent being “olatu.” It may be here apposite to make some observations on the slavery which is practised in connection with the bush-tribes of these islands. As already remarked, a wide distinction usually prevails in the Solomon Group between the inhabitants of the coast and those of the interior; and although this distinction is most evident in the case of the larger islands, it also prevails, but to a less degree, in those of smaller size. It is a noteworthy fact that the bushmen are always looked down upon by their brethren of the coast. “Man-bush” is with the latter a term of reproach, implying stupidity and crass ignorance. I have frequently heard this epithet applied to natives
- 57. who handled their canoes in an awkward manner or who stumbled in their walk whilst accompanying me in my excursions. On one occasion, when trying to obtain stone axes from the natives of Alu, I was referred with a smile to the bushmen of the neighbouring island of Bougainville, who still employ these tools. In the larger islands the bush-tribes and the coast natives wage an unceasing warfare, in which the latter are usually the aggressors and the victors—the bushmen captured during these raids either affording materials for the cannibal feast or being detained in servitude by their captors. But there prevails in the group a recognized system of slave-traffic, in which a human being becomes a marketable commodity—the equivalent being represented in goods either of native or of foreign manufacture. This custom which came under the notice of the officers of Surville’s expedition, during their visit to Port Praslin in Isabel, in 1769,[12] obtains under the same conditions at the present time. These natives were in the habit of making voyages of ten and twelve days’ duration with the object of exchanging men for “fine cloths covered with designs,” articles which were manufactured by a race of people much fairer than their own, who were in all probability the inhabitants of Ontong Java. [12] “Discoveries to the south-east of New Guinea,” by M. Fleurieu, p. 143, Eng. edit. The servitude to which the victims of this traffic are doomed is not usually an arduous one. But there is one grave contingency attached to his thraldom which must be always before the mind of the captive, however lightly his chains of service may lie upon him. When a head is required to satisfy the offended honour of a neighbouring chief, or when a life has to be sacrificed on the completion of a tambu-house or at the launching of a new war- canoe, the victim chosen is usually the man who is not a free-born native of the village. He may have been bought as a child and have lived amongst them from his boyhood up, a slave only in name, and enjoying all the rights of his fellow natives. But no feelings of compassion can save him from his doom; and the only consideration which he receives at the hands of those with whom he may have
- 58. lived on terms of equality for many years is to be found in the circumstance that he gets no warning of his fate. There are in Treasury several men and women who, originally bought as slaves from the people of Bouka and Bougainville, now enjoy apparently the same privileges and freedom of action as their fellow islanders. It is sometimes not a matter of much difficulty to single out the slaves amongst a crowd of natives. On one occasion I engaged a canoe of Faro men to take me to a distant part of their island: and very soon after we started I became aware from the cowed and sullen condition of one of the crew that he was a slave. On inquiry I learned that this man had been captured when a boy in the island of Bougainville, and I was informed that if he was to return to his native place—a bush village named Kiata—he would undoubtedly be killed. Although in fact a slave, I concluded from the bearing of the other men towards him that his bondage was not a very hard one; and he evidently appeared to enjoy most of the rights of a native of the common class. Sukai, however, for such was his name, had to make himself generally useful in the course of the day; and when at the close of the excursion we were seated inside the house of a man who provided us with a meal of boiled taro, sweet potatoes, and bananas, he was served with his repast on the beach outside. Mule, the Treasury chief, had adopted a little Bougainville bush-boy, named Sapeku, who was purchased when very young from his friends. In 1883 he was six or seven years old, and was the constant companion of the sons of the chief. He was a fat chubby little urchin, with woolly hair, and was known on board under the name of “Tubby.” His wild excitable disposition full of suspicion showed to great contrast with the calmer and more confident demeanour of his companions. He was, however, a general favourite with us, although I should add he did not possess half the pluck of his associates. Mule also possessed, at the time of our visit, a young girl, twelve or thirteen years old, who had been not long before purchased from the Bougainville natives.
- 59. I have previously referred to the existence of bushmen on some of the smaller islands. In the interior of Treasury there are a few hamlets containing each two or three families of bushmen, who live quite apart from the other natives of the island. On more than one occasion I experienced the hospitality of these bush families, who in matters of dress are even less observant than the harbour natives. They are probably the remnants of the original bushmen who occupied this island. Over our pipes, I used frequently to converse with the natives on the subject of the past history of their island; and I gleaned from them that the enterprising race at present dominant in the Bougainville Straits came originally from the islands immediately to the eastward, using Treasury as a stepping-stone to the Shortlands and Faro, and ousting or exterminating the bushmen they found in the possession of these islands. I will turn for a moment to the subject of slavery in the eastern islands of the group. In Ugi it is the practice of infanticide which has given rise to a slave-commerce regularly conducted with the natives of the interior of St. Christoval. Three-fourths of the men of this island were originally bought as youths to supply the place of the natural offspring killed in infancy. But such natives when they attain manhood virtually acquire their independence, and their original purchaser has but little control over them. On page 42, I have made further reference to this subject. Connected in the manner above shown with the subject of slavery is the practice of cannibalism. The completion of a new tambu-house is frequently celebrated among the St. Christoval natives by a cannibal feast. Residents in that part of the group tell me that if the victim is not procured in a raid amongst the neighbouring tribes of the interior, some man is usually selected from those men in the village who were originally purchased by the chief. The doomed man is not enlightened as to the fate which awaits him, and may, perhaps, have been engaged in the erection of the very building at the completion of which his life is forfeited. The late Mr. Louis Nixon,[13] one of those traders whose name should not be forgotten amongst the
- 60. pioneers who, in working for themselves, have worked indirectly for the good of their successors in the Solomon Group, once recounted to me a tragical incident of this kind on the island of Guadalcanar, of which he was an unwilling spectator. Whilst looking out of the window of his house one afternoon, he observed a native walk up to another standing close to the window and engage him in conversation. A man then stole up unperceived, and raising his heavy club above his head, struck the intended victim lifeless to the ground. Knowing too well the nature and purpose of the deed, Mr. Nixon turned away quite sickened by the sight. [13] Mr. Nixon died at Santa Anna in the end of 1882. The natives of the small island of Santa Anna enjoy the reputation of being abstainers from human flesh: but, inasmuch, as Mai the war- chief has acquired a considerable fortune, in a native’s point of view, by following the profitable calling of purveyor of human flesh to the man-eaters of the adjacent coasts of St. Christoval—a trade in which he is ably assisted by those who accompany him on his foraging expeditions—we can hardly preserve this nice distinction between the parts taken by the contractor and his customers in this extraordinary traffic. I learned from Captain Macdonald that in their abstinence from human flesh, the Santa Anna natives are not actuated by any dislike of anthropophagy in itself; but that the custom has fallen into abeyance since the chief laid the tambu-ban on human flesh several years ago, on account of a severe epidemic of sickness having followed a cannibal feast. On one occasion through the instrumentality of this resident, Lieutenant Oldham had the satisfaction of rescuing two St. Christoval natives whom Mai was carefully keeping in anticipation of the wants of the man-eaters of Cape Surville. As the result of an interview held with this chief, the two prisoners were sent on board the “Lark;” but Mai gave them up with a very bad grace, protesting that he was being robbed of his own property. It is difficult to speculate on the reflections of the victim as he lives on from day to day in constant expectation of his fate. I am told that there is a faint gleam of tender feeling shown in
- 61. the case of a man who, by long residence in the village, has almost come to be looked upon as one of themselves. He is allowed to remain in ignorance of the dreaded moment until the last: and, perhaps, he may be standing on the beach assisting in the launching of the very canoe in which he is destined to take his final journey, when suddenly he is laid hold of, and in a few moments more he is being ferried across to the man-eaters of the opposite coast. All persons whom I have met that have had a lengthened experience of the St. Christoval natives confirm these cannibal practices. They may sometimes be observed with all the horrible preliminaries which have been described in the cases of other Pacific groups; whilst, on the other hand, it may be the habit to purchase and partake of human flesh as an extra dainty in the daily fare. Captain Redlich, master of the schooner “Franz,” who visited Makira on the south side of St. Christoval in 1872, states that he found a dead body in a war-canoe dressed and cooked whole. He was informed by Mr. Perry, a resident, that he had seen as many as twenty bodies lying on the beach dressed and cooked.[14] In 1865, Mr. Brenchley noticed at Wano, on the north coast of this island, the skulls of twenty-five bushmen hanging up under the roof of the tambu-house, all of which showed the effects of the tomahawk and all had been eaten.[15] At the present time it is not an easy matter for any person not resident in the group to obtain ocular evidence of cannibalism, since the natives have become aware of the white man’s aversion to the custom. I have, however, frequently seen the arm and leg bones of the victim consumed at the opening of a new tambu-house, as they are usually hung up over the entrance or in some other part of the building. The natives, however, are generally reluctant to talk much about these matters; and I believe the residents, in such matters, prefer to trust more to the testimony of their own eyes than to the statements of the natives. [14] Journal of the Royal Geographical Society for 1874 (vol. 44), p. 31. [15] “Cruise of H.M.S. ‘Curaçoa,’” by J. L. Brenchley.
- 62. Welcome to our website – the ideal destination for book lovers and knowledge seekers. With a mission to inspire endlessly, we offer a vast collection of books, ranging from classic literary works to specialized publications, self-development books, and children's literature. Each book is a new journey of discovery, expanding knowledge and enriching the soul of the reade Our website is not just a platform for buying books, but a bridge connecting readers to the timeless values of culture and wisdom. With an elegant, user-friendly interface and an intelligent search system, we are committed to providing a quick and convenient shopping experience. Additionally, our special promotions and home delivery services ensure that you save time and fully enjoy the joy of reading. Let us accompany you on the journey of exploring knowledge and personal growth! textbookfull.com