Pitch detection from singing voice, advantages, limitations and applications of pitch in voice processing

IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308
_______________________________________________________________________________________
Volume: 04 Issue: 04 | Apr-2015, Available @ http://www.ijret.org 616
PITCH DETECTION FROM SINGING VOICE, ADVANTAGES,
LIMITATIONS AND APPLICATIONS OF PITCH IN VOICE
PROCESSING
Sumeet S. Andhalkar1
1
Student, Department of Computer Engineering, G.H.R.C.E.M., Wagholi, Pune, Maharashtra, India
Abstract
Detecting the Pitch of singing voice in the presence of music is a challenging work to do. In this paper we study the problem of
finding pitch range or value by analyzing the voice. We also review the vibratos and tremolos used to detect the singing voice. We
just take review of detecting pitch from musical record and merits and demerits of pitch. Pitch is very important characteristic of
voice and by detection it we can develop other different research works in voice processing field. We also study the advantages of
pitch and its limitations. We can also study the future application that we can develop by using pitch. Pitch can be defined as the
extent to which sound is high or low. It is the level of sound which may be high or low. It is also known as particular level of
intensity of sound. Pitch detection is known as determining the level of intensity of voice. By using pitch we can detect various
other characteristics of voice. We can develop many applications those are very useful for the benefit of human being or mankind.
Pitch detection is crucial task in singing voice separation also. Pitch detection also play important role in Musical information
retrieval, Identification of the singer and in lyric recognition. Pitch can identify gender of singing voice. It means it can identify
the gender of the singer. Pitch also can examine or find the time of voice recording or the time slot of voice recording. So it is very
much necessary to study pitch and its future work.
Keywords: Pitch, Sinusoidal, Tremolos, Vibratos, Timbre.
--------------------------------------------------------------------***----------------------------------------------------------------------
1. INTRODUCTION
The Voice Processing is the study of voice signals and the
processing approaches of these signals. The signals are
usually treated in a digital depiction, so voice processing can
be viewed as a extraordinary case of digital signal
processing.
Pitch can be defined as the extent to which sound is high or
low. It is the level of sound which may be high or low. It is
also known as particular level of intensity of sound. Pitch
detection is known as determining the level of intensity of
voice. Pitch detection is determining the extent of voice to
which it is high or low. Pitch detection is very important in
some related tasks of voice processing. Pitch detection is
crucial task in singing voice separation also. Pitch detection
also play important role in Musical information retrieval,
Identification of the singer and in lyric recognition. Pitch
can identify gender of singing voice. It means it can identify
the gender of the singer. Pitch also can examine or find the
time of voice recording or the time slot of voice recording.
One method to find pitch is extract melody from the music
record. Then find the pitch of that particular singing voice.
Alternative way suggests separation of voice and music and
then only find sinusoidal wave of singing voice. Then
extract necessary data from wave and then we can estimate
the pitch.
The pitch ranges are usually large to cover most of the
possible pitches of singing voice such as from 80 Hz to 800
Hz. However, it is unlikely that pitch changes in such a wide
range in a short period of time. Furthermore, the upper pitch
boundary of singing can be as high as 1400 Hz for sharp
singers.
The detection of the pitch of a harmonic sound in the
presence of other sounds has been mainly studied in the
context of speech segregation, speech enhancement and
other related fields. As per the study, before applying
methods developed for pitch detection for speech, it is
helpful to make a comparison between singing and speech.
We can develop large number of applications in computer
technology by using this pitch range. These applications
may be useful in technology, medical science, security
applications etc. which are enlisted in this paper. So pitch is
a crucial factor in voice processing field. So in later part of
the paper we will study the method to estimate the pitch and
advantages and disadvantages and applications and future
works of pitch and pitch detection.
1.1 Need of the Work
The research in this area is required to develop many
applications related to the work. Many applications can be
created for the enrichment of mankind.
The applications like decease identification gender
identification or for some security purposes we can use these
applications.
Timbre – Timbre is very special character of voice. It
identifies an object by its sound or voice. Timbre is special

_______________________________________________________________________________________
thing that separated or distinguish two or many sounds from
each other. This timbre can be computed by using pitch
factor.
2. PITCH DETECTION
We can detect trend of the pitch by Trend estimation
algorithm. There are basically four steps.
1] Vocal Improvement.
2] Sinusoidal Mining.
3] Instrumental Pruning.
4] Pitch estimation.
2.1 Vocal Improvement
Here, The Vocal improvement methods are used to improve
the voice from musical record. It also increases the quality of
the musical record. They use Harmonics and percussive
sound separation. Harmonic instruments are like flute,
Guitar. Percussive instruments are like Tabla, Drum. These
methods reduce the energy of harmonic instrument and
reduce the energy of percussive instruments. This means that
the signals of these harmonic and percussive devices are
made weaker. So that we can easily catch voice signal more
rich and loud and clean.
Here, we use first step only. The signals of harmonic
instruments are reduced only. The Percussive instruments do
not make any difficulty to sense the voice and its pitch. So,
this is the first step in Pitch Detection.
2.2 Sinusoidal Mining
The sinusoidal partials can be mined by using this step. The
Waves of the musical record and the vocal and musical parts
can be found in this step. Some defective parts or peaks
from these waves can be detached. These peaks do not
belong to the musical file. Then, regular peaks can be used
to form Sinusoidal partials. After peak selection one
grouping algorithm is used to reduce the gaps.
It consists of three steps: early grouping, re-grouping and
purifying.
i) Early grouping: It starts by selecting any ungrouped peak
as the first peak in the group and recursively groups other
peaks neighboring to the group until each peak fits to a
group. The closest peak is selected if the difference between
it and the previous peak is not larger than a semitone. If no
peak exists, the group is divided into two.
ii) Re-grouping: The objective of this step is to unite the
partials that are possible to be originated from the same.
iii) Purifying: Here study says that we require filling the gap
in between the partials. This gap filling is the basic concept
of purifying. Because the re-grouping step introduces gaps
in a group, this step chooses one peak from overlapping
peaks and fills the gaps for each partial. Thus we can purify
the signal. For filling the gaps, values of frequency and
magnitude are interpolated according to the adjacent peaks.
As we fill all the gaps, then we can say we have purified or
enhanced these partials effectively.
2.3 Instrumental Pruning
First we should know the actual meanings of some technical
words and then we can start up with pruning.
Vibrato-Vibratos are the vocal partials.
Tremolos-Tremolos are the instrumental partials.
This Pruning step categorizes vibratos and tremolos. This
means it simply classifies the vocal partials and instrumental
partials. As vibratos are vocal partials, Vibratos refer to the
variation of pitch. As Tremolos are instrumental partials,
Tremolo refers to the periodic vibration of intensity. As per
the study and survey, this whole step used to detect singing
voice. Two qualities are computed to describe vibrato and
tremolo the rate and the extent of vibrato or tremolo.
According to study and literature, for human singing voice,
the average rate is around 6Hz for both vibrato and tremolo.
Hence we determine the relative extent around 6Hz by using
the Fourier transform for both vibrato and tremolo.
2.4 Pitch Estimation
According to literature study, some graphs are obtained by
the above step. By analyzing those graphs the series or
variety of the singing voice can be estimated. This is an
important step for actual picking of range of the pitch
variation. Pitch varies as singer proceeds with the song. This
extent to which the voice is high or low as singer sings a
song is pitch variety or range. This is the basic thing to
study. If we can get the pitch range then we can develop
variety of applications based on voice processing. By
analyzing those graphs and some figures we can easily
estimate the range of the pitch.
The pitch obtained in this process/method is not a value, but
it a range or variety of values that can be used for further
processes. The following/ below model can be used to
extract these pitch ranges from vocal partials.
Fig 1 Steps in Pitch Detection

_______________________________________________________________________________________
3. MERITS OF PITCH DETECTION.
There are many advantages of pitch. By detection of pitch
we can enhance the research area and enrich research in
voice processing. The advantages of the pitch are elaborated
below.
i) Pitch is the most basic characteristic of voice. So by
detecting the pitch we can do many other things like
intensity and timbre detection.
ii) Separation of voice and music - If we have the pitch
then we can separate the human voice and musical parts. We
have to decide pitch range for each alphabet or for each
word. Then by analyzing various pitch ranges from the
musical record we can decide the words and separate them
out to the paper.
iii) Gender Identification - As we have the pitch, we can
easily detect the gender of the singer as male and female
voices have different pitch range. As male and female pitch
ranges or levels are different as differentiable, then we can
definitely identify the gender of
iv) Lyric Separation/Extraction- We can separate out the
words of the song sung by the singer. Pitch is an important
aspect here. Lyric is the words spoken by singer or words of
the song. These words have different pitch ranges. So by
detecting pitch, we can easily detect the of the song. So we
can separate out words or lyric from the record.
v) Information Retrieval- We can retrieve large amount of
information from these musical records, if we have the
pitch. Various instruments have various pitch ranges. So
they can give more relative musical information. Large
amount of characteristics can be retrieved from music or
song. Pitch, timbre and other characters of voice cn be
retrieved from the voice easily by using pitch.
vi) Instrument Detection: Various instruments have
various pitch ranges. Vocal partials are known as vibratos
whereas Musical partials are known as Tremolos. There are
two types of musical instruments, first one is harmonic
instrument and second one is percussive. Harmonic
instruments are mostly key pressing or tense wired
instruments those includes flute, guitar. Percussive
instruments includes tabla, drum etc. So, these instruments
have different pitch range, by estimating pitch we can
identify the instrument easily.
vii) Raag Identification- Identifying Raag is definitely
beneficial for new classical singers or classical music
students. As Raag is digitized here, they can be stored and
easily studied by the students.
vii) Identification of deceases- We can identify deceases
related to the vocal system like Lungs, throat, other organs
included in production of voice. We also can determine the
problems related breathing organs and some vocal organs.
This research will give vital benefit to medical science
research and will work for benefit of mankind.
4. LIMITATIONS OF PITCH DETECTION
There are some limitations also which are discussed below.
4.1 Complexity
The entire process of detection of pitch is very complex and
difficult to implement. The method requires some electronic
devices and accurate software. It is very difficult to develop
software which is very complex.
4.2 Accuracy
Any kind of mistake or wrong conversions leads to an error.
These are not allowed. Any wrong conversion or input may
change entire output and system may work wrongly.
Accuracy level must be very high. Any mistake or bug or
error can lead to failure of application or will give wrong
output.
4.4 Separation and comparison:
The separation of voice, noise, music, words, vocal parts are
difficult to implement. Here, high level of accuracy
required. Then to compare that data and to produce result
more complex methods are required. So, this comparison
must not be error prone.
5. APPLICATIONS OF PITCH DETECTION
By extracting pitch range or trend or variety we can asses
many things. By using or analyzing those things we can
develop many soft wares and applications those may add
value to the society.
i) Singer Identification System: We can identify singer by
estimating pitch range of his voice. This application can be
used for various purposes related to music. We can identify
the singer by using pitch variety in his singing. The pitch
range can identify the singer who is performing. This can be
done at real time also if system have the historic samples of
singer‟s voice.
ii) Confirmation of voice: Many times controversies related
to recorded voice, happening in real time world. These
controversies can be resolved by using voice identification
system. This system will compare person‟s voice with
recorded voice and check for similarities.
iii) Voice Separation System: Separation of voice from
musical record is required in special cases. To make only
music file, the vocal parts are removed. Karawaks are made
by using this technique.
iv) Lyric Identification System: This system identifies
words from the song or speech or voice. The Lyric are
separated from the record by using pitch range. As each
word has its own pitch range.
v) Gender Identification System: By using pitch we can
identify the gender of singer or person. Male and female
voices have different pitch ranges. So by comparing these
ranges we can determine the gender from voice of the
person.

_______________________________________________________________________________________
vi) Raag Identification System: As pitch is basic
characteristic of voice, by using pitch we can estimate raag
from the classical music or classical song. Raag‟ detection is
also needful when application deals with classical music.
Raag is central structure in Indian classical music. Raag is
also known as unique set of complex melodic gestures. In
one song similar bits and trend is repeated constantly and
continuously. That repeated pattern is known as „Raag‟.
Same pattern of wave forms are repeated that can be
collected and stored to identify the raag. If we can separate
that repeated pattern then we can easily separate, identify
and store that pattern as raag.
vii) Decease detection system: Various deceases related to
the lung and breathing system can be identified by using
voice processing techniques. As lungs and entire breathing
system works to produce voice, by detecting and analyzing
various aspects of voice we can detect deceases related to
breathing system of human body.
vii) Composing Song: By analyzing and understanding
digital raag, music composer can compose many songs by
using same raag. This is more important advantage.
vii) Identification of Time of Recording of Voice: We can
identify the time slot in which the voice is recorded. The
pitch and timbre combination can lead to identification of
recording time. The system can identify time slot of
recording like morning, evening or afternoon recording.
viii) Age Identification System- Age of the singer can be
assessed by using this pitch and timbre. Age of singer or age
group can be identified by the software using pitch and
timbre values.
ix) Security Applications- We can create various devices
for authentication of person, similar to biometric devices.
We can identify suspected enemy by his voice. We can use
these records as evidence in judicial matters. These are
definitely the future scopes that can benefit mankind.
6. CONCLUSION
This paper studies method for singing pitch detection in
musical recordings. This paper studies method of detecting
pitch range from voice. The studied method says that by
following some steps it can extract pitch from voice. We
studied that Pitch can be defined as the extent to which
sound is high or low. It is the level of sound which may be
high or low. It is also known as particular level of intensity
of sound. Pitch detection is known as determining the level
of intensity of voice. Pitch detection is determining the
extent of voice to which it is high or low. This paper also
studies advantages and disadvantages of the pitch. This
paper also studies applications and future scope of the pitch
i.e. uses of pitch. We mainly studied that there is a method
that can detect pitch from singing voice and we can obtain
its range by the respective method.
ACKNOWLEDGEMENTS
The Singing voice pitch detection and voice separation and
supporting documents are the work of many people,
including Chao Ling Hsu, DeLiang Wang, Jyh-Shig Roger
Jang. I also thank teachers and authors of different papers
which are referred for study. I also thank my project guide
and my teachers who always guided me. I also thank my
classmates and friends who helped me in my study.
REFERENCES
[1]. Chao Ling Hsu, Deling Wang,Jhy Shing”A Trend
Estimation Algorithm For Singing Pitch Detection In
Musical Recordings”.
[2]. Chao Ling Hsu, Deling Wang”A Tandem Algorithm
For Singing Pitch Extraction and Voice Separation From
Musical Accompaniment”.
[3]. L. Regnier, G. Peeters “singing voice detection in music
tracks using direct voice vibrato detection”.
[4]. Yipeng Li and DeLiang Wang ”detecting pitch of
singing voice in polyphonic audio”.
[5]. L. Regnier, G. Peeters, ”singing voice detection in
music tracks using direct voice vibrato detection”.
BIOGRAPHIES
Andhalkar Sumeet Sudhakar, Studying
Masters of Engineering (Computer
Engineering) in G.H.R.C.E.M., Wagholi,
Pune. Has completed Bachelor of
Engineering in Information Technology
from B.V.C.O.E. Kolhapur

Pitch detection from singing voice, advantages, limitations and applications of pitch in voice processing

More Related Content

What's hot (14)

Viewers also liked (9)

Similar to Pitch detection from singing voice, advantages, limitations and applications of pitch in voice processing (20)

More from eSAT Journals (20)

Recently uploaded (20)

Pitch detection from singing voice, advantages, limitations and applications of pitch in voice processing