Snorm–A Prototype for Increasing Audio File Stepwise Normalization

1
Sachin M. Narangale, 2
Prof. Dr. G. N. Shinde. Int. Journal of Engineering Research and Application
www.ijera.com
ISSN : 2248-9622, Vol. 6, Issue 3, ( Part -4) March 2016, pp.171-174
www.ijera.com 171|P a g e
Snorm–A Prototype for Increasing Audio File Stepwise
Normalization
1
Prof. Dr. G. N. Shinde
1
School of Media Studies, Swami Ramanand Teerth Marathwada University, Nanded-431606, India
2
Pro-Vice Chancellor, Swami Ramanand Teerth Marathwada University, Nanded-431606, India
ABSTRACT
This paper introduces a novel concept SNORM (Step NORMalization) for increasing normalization. It is a
prototype algorithm for increasing normalization based on loudness factor of the audio. The function effect
normalization plays a vital role in loudness control. The proposed experiment carried out for increasing
normalization based on step wise increase yield the variations in peaks of the audio file. The experimental
results are shown in the form of graphical analysis of the plot spectrum values of frequency analysis. From the
results, it is clearly apparent that, the normalization values are increased at different levels. The function
SNORM can set a new benchmark in the field of audio industry for the processes of increasing normalization.
SNORM can be substantial in the audio broadcast systems for applications in live audio streaming, news
broadcast, sports coverage, live programming where the loudness control mechanism is essential. For the
selective or predictive loudness control systems SNORM can be effectively applied.
Keywords: Normalization, Loudness Control, Track normalization, Audacity, Audio Retrieval, Text to Speech
application, Frequency Analysis
I. INTRODUCTION
The retrieval systems for audio are major
game players in the music industry. The content-
based audio information retrieval system (AIRS)
for different electronic equipments is creating lot of
attention. AIRS defined properly will definitely
perform various retrieval operations including
recognition, referencing and recommendation.
Content-based music information retrieval (MIR)
systems such as Shazam, SoundHound, and
Gracenote have already been developed for the
iPhone, iPad, and other similar Smartphone devices
[1]. Audio retrieval is a mechanism that searches
music that is being played. The audio identification
and retrieval system has to identify the different
background noises. For building AIRS, audio
tokenization or unique identification is necessary.
These unique identification measures act as the
identity of the audio to be retrieved from the
database. This audio unique identification measure
contains brief details of the audio file or sometimes
a frame of audio too. The growth of the audio
industry has invariably led to the demand for quick
and correct audio data retrieval. The audio retrieval
system resulting lot of information none of the use
is more. The audio retrieval rates need to be
improved. To improve the retrieval rate, before the
query is to be fired for the audio, first, it is essential
to create a audio lookup reference table (ALURT).
For audio to be matched the audio unique
identification is maintained as the private key of
retrieval. Due to this the audio retrieval probability
of intended audio increases quantitatively. In this
research, a mechanism in the form of prototype
pre-processing method for retrieving the intended
audio from ALURT is presented. The novel
proposed mechanism SNORM works on the
principle of normalization. A study has been
evaluated for increasing normalization stepwise to
counter the normalization values of the audio file.
Normalization is a process of reducing
peak amplitude of the audio signal to a defined
intended level or to the corresponding average of
the frame of an audio signal. Normalization is the
process of modification of the amount of gain or
amplitude of the audio signal. The peak amplitude
is reduced to a target level in normalization. Audio
Normalization is a pre-process of the system used
in audio compression [2]. Another effective use of
normalization can be seen in the process of
loudness control. Normalization works out to
smooth out the variations in the loudness. Peak
control and loudness control can be achieved using
normalization [3].
Normalization is an effect provided in the
Audacity to guarantee the complete representation
of every audio feature element. The process
involves the methodology of subtracting the mean
of audio feature and the resultant value is divided
by its standard deviation [4].
Various application areas of the
normalization process are:
• Process of loudness control
• Smooth out the variations in the loudness
RESEARCH ARTICLE OPEN ACCESS

1
www.ijera.com
• To guarantee the complete representation of
every audio feature element
• To check audio compatibility with the
reference signal
• For every audio input track, the speaker may
not speak at an expected level
• In the research of text to speech mechanism,
audio has to be normalized
• Different channel surfing over television
network if doesn’t have audio normalization
done, the audience may get disturbed by the
sudden increase in loudness.
II. RELATED WORK
Normalization has been carried out on
attacked audio samples before the watermark
extraction to be compatible with the reference
signal [5]. Reference signals are the audio reference
signals which can be considered while ALURT
preparation. For the listening tests, the
normalization of the gains of the audio is
necessary. The auralized samples undergo
normalization process on the equivalent level for
the listening test by altering gains. The increase in
graphical computing power and the system
capacity is demanding increased aspiration to
achieve higher visual reliability in non-real world
environments and the video games. As the visual
fidelity perception is getting more attention, the
role of audio in multimedia is getting
inattentiveness. But the question is, whether the
visual fidelity is sufficient for higher reliability?
No, the audio perception is equally important in the
reliability. To achieve the attention in the visual
fidelity, audio needs to be normalized at particular
levels [6]. For live broadcast systems, for example
news broadcast, every time the speaker may not be
a perfect speaker with monotonous maintained
audio levels. Sometimes, the environmental factors
may generate some noise that increases the audio
signals all of a sudden, disturbing the continuity of
the viewer. Similarly, in case of different channel
viewing over television network, some channel
may not control the loudness of the audio signals.
The audio loudness control mechanism through
normalization is essential.
In the case of text-to-speech paradigm, the
text normalization based on textual grammar and
pronunciation is needed. The audio generation
based on the text reading has to have the audio
normalization pre-processing methodology for
loudness control [7].
The classification systems to become
robust to loudness and channel changes like CD
channel to telephone channel use normalization [8].
III. EXPERIMENTAL SETUP
This research includes various
experiments executed on open source freeware
Ubuntu, a Linux desktop operating system. For
audio recording and editing an open source, free
and cross-platform software Audacity was used.
Audacity gets perfectly installed and configured
with the Ubuntu. The features of Audacity in this
research are used for prototype designing related to
normalization. For GUI between Audacity and
Ubuntu a C++ library wxWidgets was used.
wxWidgets facilitates developers with a single
code base for creating cross-platform applications
for Windows, Mac OS X, Linux and other
platforms.
A desktop system with hardware
configuration 2GB RAM, i3 processor, 3.08GHz
and 350GB HDD was used for experiments.
Experiments started with recording a sample audio
file “hello123.mp3” using Audacity recording
feature. Normalization experiments were conducted
on “hello123.mp3”. Different stepwise
normalization parameters were incorporated in the
code file “Normalize.cpp”. The frequency analysis
spectrum plots were captured using screen
capturing tools and presented in following figures.
Plot spectrum analysis values were analyzed
against the frequency values in squared form. The
comparative analysis chart has been presented in
this paper below.
Snorm Algorithm
Step 1: Start
Step 2: ProcessFirst – take a track
Step 3: Transform track into bunch of buffer blocks
Step 4: Get the offset value and ratio parameters
set.
Step 5: SNORM function
Step 6: SNORM Step 2,4
Step 7: Adjust frames based on step parameters and
offset values.
Step 8: Analyze Track
Step 9: Analyze Data
Step 10: Return
IV. RESULTS AND DISCUSSION
This research presents the results of
executing SNORM stepwise function for increasing
gain of audio file by molding the effect
normalization in Audacity. For analyzing the
results the graphical chart analysis has been used.
The representation of increased gain with SNORM
has been depicted in the figure 5.
Figure (1) is the graphical Plot spectrum
of the audio file “hello123.mp3” without any
editions to it. The frequency analysis plot spectrum

1
www.ijera.com
has been shown in this figure. The audio file
“hello123.mp3” is a sample audio file created using
recording feature of Audacity.
Figure 1: Plot spectrum of the Original audio file
“hello123.mp3” – frequency analysis
of the audio file “hello123.mp3” after the
normalization process has altered the gain of the
audio file. Normalization as stated earlier, is a
process of altering peak values.
Figure 2: Plot spectrum of the Normalized audio
file “hello123.mp3” – frequency analysis
Figure (3) is the graphical Plot spectrum of the
audio file “hello123.mp3” after the normalization
process has altered the gain of the audio file up to
step 2. SNORM is executed on the audio file. This
figure, clearly it is seen that, the gain values has
been altered. The amplitude values are changed.
The SNORM has increased the normalization of
audio files.
Figure 5 Graphical Analysis of Original audio file
and SNORM applied files
Figure 3: Plot spectrum of the SNORM step 2
Normalized audio file “hello123.mp3” – frequency
analysis
of the audio file “hello123.mp3” after the
normalization process has altered the gain of the
audio file up to step 4. SNORM is executed on the
audio file. This figure, clearly it is seen that, the
gain values has been altered. The amplitude values
are changed. The SNORM has increased the
normalization of audio files. The increase in the
peaks or amplitude is clearly seen. The objective of
step 4 normalization is achieved. The frequency
analysis can be identified as the symmetrical
representation in this figure.
Figure 4: Plot spectrum of the SNORM step 4
Normalized audio file “hello123.mp3” – frequency
analysis

1
www.ijera.com
Figure (5) is the graphical waveform presentation
of the original audio file “hello123.mp3” and after
the normalization process has altered the gain of
the audio file up to step 4. The proposed prototype
SNORM with step 2 and step 4 are resulting
increase in the normalization values of the audio at
an average of 101 and 98 percentages respectively.
V. CONCLUSION
This research emphasizes the concept of
increasing normalization for smoothing out the
variations in gain or loudness. The function effect
normalization plays a vital role in loudness control.
The proposed experiments carried out for
increasing normalization based on step wise
increase yield the variations in peaks of the audio
file. SNORM is a novel prototype algorithm for
increased normalization based on loudness factor
of the audio. The proposed prototype SNORM with
step 2 and step 4 are resulting increase in the
normalization values of the audio at an average of
101 and 98 percentages respectively. SNORM can
be substantial in the audio broadcast systems for
applications in live audio streaming, news
broadcast, sports coverage, live programming
where the loudness control mechanism is essential.
For the selective or predictive loudness control
systems SNORM can be effectively applied.
REFERENCES
[1]. Dae-Jin Kim, Ddeo-Ol-Ra Koo, “Analysis
of Pre-Processing Methods for Music
Information Retrieval in Noisy
Environments using Mobile Devices”,
International Journal of Contents, Vol.8,
No.2, Jun 2012
[2]. Zainab T. Drweesh, Loay E. George,
“Audio Compression Based on Discrete
Cosine Transform, Run Length and High
Order Shift Encoding”, International
Journal of Engineering and Innovative
Technology (IJEIT), ISSN: 2277-3754,
ISO 9001:2008 Certified, Volume 4, Issue
1, July 2014
[3]. Manik Gupta, Jozsef Pinter, “Loudness
Measurement and Control”, Proceedings
of NAB Broadcast Engineering
Conference, 2012, pp.195-200
[4]. M. Liu, C. Wan, L. Wang, “Content-based
audio classification and retrieval using a
fuzzy logic system: towards multimedia
search engines”, Soft Computing 6 (2002)
357 – 364 Springer-Verlag, 2002,
10.1007/ s00500-002-0189-3, pp.357-364
[5]. Janusz Cichowski, Andrzej Czyżewski,
Bożena Kostek, “Analysis of impact of
audio modifications on the robustness of
watermark for non-blind architecture,
Multimed Tools Appl (2015) 74:4415–
4435 Springerlink.com
[6]. David Rojas, Bill Kapralos, Andrew
Hogue, Karen Collins, Lennart Nacke,
Sayra Cristancho, Cristina Conati, Adam
Dubrowski, “The Effect of Sound on
Visual Fidelity Perception in Stereoscopic
3-D”, IEEE Transactions On Cybernetics,
2013, pp.1-12
[7]. Arun Soman, Sachin Kumar S., Hemanth
V. K., M. Sabarimalai Manikandan, K. P.
Soman, “Corpus Driven Malayalam Text-
to-Speech Synthesis for Interactive Voice
Response System”, International Journal
of Computer Applications (0975 – 8887),
Volume 29– No.4, September 2011,
pp.41-46
[8]. Hadi Harb, Liming Chen, “A general
audio classifier based on human
perception motivated model”, Multimed
Tools Appl (2007) 34:375–395 DOI
10.1007/s11042-007-0108-9 Springer
Science

Snorm–A Prototype for Increasing Audio File Stepwise Normalization

More Related Content

What's hot (17)

Viewers also liked (20)

Similar to Snorm–A Prototype for Increasing Audio File Stepwise Normalization (20)

Recently uploaded (20)

Snorm–A Prototype for Increasing Audio File Stepwise Normalization