Speech compression analysis using matlab

IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308
__________________________________________________________________________________________
Volume: 03 Issue: 01 | Jan-2014, Available @ http://www.ijret.org 538
SPEECH COMPRESSION ANALYSIS USING MATLAB
Manas Arora1
, Neha Maurya2
, Poonam Pathak3
, Vartika Singh4
1, 2
M.Tech Student, Digital Communication, BBDNIIT, Uttar Pradesh, India,
3
Associate Professor, 4
Sr. Lecturer, Electronics & Comm., BBDNIIT, Uttar Pradesh, India
Abstract
The growth of the cellular technology and wireless networks all over the world has increased the demand for digital information by
manifold. This massive demand poses difficulties for handling huge amounts of data that need to be stored and transferred. To
overcome this problem we can compress the information by removing the redundancies present in it. Redundancies are the major
source of generating errors and noisy signals. Coding in MATLAB helps in analyzing compression of speech signals with varying bit
rate and remove errors and noisy signals from the speech signals. Speech signal’s bit rate can also be reduced to remove error and
noisy signals which is suitable for remote broadcast lines, studio links, satellite transmission of high quality audio and voice over
internet This paper focuses on speech compression process and its analysis through MATLAB by which processed speech signal can
be heard with clarity and in noiseless mode at the receiver end .
Keywords: Speech compression, bit rate, filter, MPEG, DCT.
----------------------------------------------------------------------***----------------------------------------------------------------------
1. INTRODUCTION
Data compression is a technique in which data content of the
input signal to system is compressed so that original signal is
obtained as output and unwanted or undesired signals are
removed. Therefore when speech signals are used in the form
of data it is termed as SPEECH COMPRESSION.
Speech is a very basic way for humans to convey information
to one another. Speech has a small bandwidth of 4 kHz. Speech
compression involves coding of real-time audio signals at the
lowest possible bit rates. The compression of speech signals has
many practical applications. It is used in digital cellular
technology where many users share the same frequency
bandwidth. Compression allows more users to share the system
at a particular time. It can also be used for digital voice storage
that are used for answering machines and pre-recorded
telephone calls that are used for purpose of providing any kind
of information to user or advertising. For a given memory size,
compression allows longer messages to be stored than
otherwise [1].
2. SPEECH COMPRESSION
Speech compression enables efficient storage and transmission
of data. There may be varying amounts of compression in data
according to the sampling rate used. This gives different levels
of system complexity and compressed quality of speech data.
The recorded waveform which is compressed can be
transmitted with or without loss. Therefore, there are two types
of compression: lossy and lossless. The digital audio data is
processed through mixing, filtering and equalization. The
speech signal is fed into an encoder that uses fewer bits than
original audio data bit rate. This results in reducing the
transmission bandwidth of digital audio streams and also
reduces storage size of audio files. [15] Compression may be
lossy or lossless. Lossy compression is transparent to human
audibility but lossless being have a compressing factor from 6
to 1. [15].An uncoded speech signal is [17]
Fig 1: Uncoded speech signal
At the decoder, bit stream decoding bit is done which is
followed by frequency sample reconstruction. Finally
frequency-to-time mapping produces decoded speech signal.
DCT is used to compress speech signals. [15].
Discrete Cosine Transform (DCT) is a main transform or
mapping method used which maps the time domain into
frequency domain. A DCT expresses a finite sequence of data
points in terms of a sum of cosine functions oscillating at
different frequencies. The discrete cosine transform is a linear,

__________________________________________________________________________________________
invertible function defined as F: RN RN (where R denotes
the set of real numbers), or equivalently an
invertible N × N square matrix. Mathematically DCT is
calculated from (1) [2]
1,.......0
2
1
cos
1
0




















 


Nk
N
kn
xX
N
n
nk
(1)
The cosine graphs define the nature of frequency used in DCT
transform and define the frequency of the speech signal. The
cosine graphs define the window function. The window
function has cosine function as (2).
(2)
Basic cosine graphs used in the discrete cosine transform are:
Fig 2: Cosine graphs used in DCT
The speech compression using DCT method is an efficient
method by which we can modify the frequency of the signal
and analyze the signals in different frequency ranges in a fixed
frequency domain. DCT also effectively compresses the speech
signals and maintain the audibility feature of speech signals.
3. NEED OF SPEECH COMPRESSION
A basic question arises what is the need of speech
compression? Answer lies within the requirement of
appropriate bit rate for high quality speech signal. Bit rate is
defined as number of sample per second which is given by
Rbits calculates from (3) where fs is the sampling frequency
and n describes the number of bits.
nfR Sbits  (3)
The requirement of achieving high bit rate and clarity of audio
signals basically defines the need of speech compression[14]. A
given frequency bandwidth can be used by a number of users at
a time and moreover, larger size speech data files can be stored.
4. SPEECH COMPRESSION TECHNIQUE IN
MPEG TECHNOLOGY
A speech signal is a complex signal which has a wide spectrum
and MPEG helps in perceptual phenomena for our ears. The
large frequencies and smaller frequencies below masking
threshold are inaudible to human ears. MPEG standard helps in
defining standard and filtering audio signals in the available
bandwidth and maintains signal to quantization ratio. In
MPEG-1 audio compression is performed as given in figure 1
and involves two process. Filter bank which uses filters divides
spectrum of incoming signals in sub bands. The quantizer
quantizes the sub bands. [10]
Fig 3: Block Diagram of MPEG-1 Encoder
The basic method by which speech compression in MPEG
technology is done is by utilizing an MPEG encoder and
decompression is performed by MPEG decoder. Figure 3 shows
block diagrams of the MPEG/speech encoder and decoder. [11,
12]
In MPEG, compression of the input stream passes through a
filter bank which divides the input into multiple sub bands. The
input is simultaneously also passed through a psychoacoustic
model which is used to determine the signal-to-mask ratio of
each sub band. The bit or noise allocation block utilizes the
output of psychoacoustic model and a decision is made about

__________________________________________________________________________________________
the total number of code bits which are available for
quantization process [3, 13].
Code bits also help in removing quantization noise. The output
of compressor is quantized audio samples and these formats of
the data are transformed into a decodable bit stream [3, 13].
In MPEG, decompression of the speech signals simply reverses
the formatting and then reconstruction of quantized code bits is
done then reconstructs the quantized sub band values, which
are finally transformed into the set of sub band values into a
time-domain audio signal [3, 13].
SPEECH ENCODER
Fig 4: Speech Encoder block diagram
SPEECH DECODER
Fig 5: Speech Decoder block diagram
The MPEG speech standard has three distinct layers for
compression.
 Layer I is the basic algorithm
 Layers II and III are enhancements of Layer I. Each
successive layer improves the compression performance
at the increasing cost of encoder and decoder complexity
[14].
4.1 Layer I:
The Layer I algorithm uses the basic filter bank which is found
in all layers. This filter bank divides the audio signal into 32
constant-width frequency bands. The filters are simple and
provide time and frequency resolutions by which it is easily
perceived by human ears. The design of Layer I has three
concessions. First, the 32 constant width bands which do not
accurately reflect the ear’s critical frequency bands. Second, the
filter bank and its inverse are not lossless transformations.
Thirdly, adjacent filter bands have a significant frequency
overlap [4, 5, 6, 7, 8, 15]
4.2 Layer II:
The Layer II algorithm is a simple enhancement of Layer I. It
improves compression performance by coding data in larger
groups. By the Layer II encoder forms frames of 3 by 12 by 32
that is total of 1152 samples per audio channel as compared to
Layer I codes data in single groups of 12 samples for each sub
band .Layer II removes stereo redundancy coding and there is
one bit allocation The encoder encodes with a unique scale
factor for each group of 12 samples to avoid audible distortion.
The Layer II algorithm improves performance as it uses
efficient code for representing the bit allocation, the scale factor
values, and the quantized samples. [4, 5, 6, 7, 8, 15]
4.3 Layer III:
The Layer III algorithm is a much more refined
approach.[16,17] though it is based on the same filter bank
found in Layers I and II. Layer III compensates for some filter
bank deficiencies by processing the filter outputs with a
modified discrete cosine transform (MDCT) [4, 5, 6, 7, 8, 15].
These all the three layers completely design the methods of
audio compression in MPEG technology.
5. IMPLEMENTATION IN MATLAB
MATLAB is a useful tool which is used to analyze speech
signals which are read in >wav format. Following commands
are used to analyze.
 wavread: it reads speech signal
 windowsize: defines window function of transformation
 wavplay: it produces speech signal after transformation
 length: defines length of speech to be processed by
transforming principle
 dct: performs discrete cosine transform
 idct: performs inverse discrete cosine transform
6. OUTPUT IN MATLAB
In speech signal analyzing through MATLAB we obtain
following spectrums
 Speech Signals with different Amplitudes(as input data)
 Portion of signal according to length and window-size
 Speech spectrograms

__________________________________________________________________________________________
Fig 6: Speech Signals with different Amplitudes
Fig 7: Portion of speech signal to be processed
Fig8: Speech Spectrograms
7. PRESENT SCENARIO
MPEG technology is an ever developing technology in the field
of communication in which speech compression is finding its
space and utility. Currently speech compression is used in
satellite communications, DBS TV’s, TATA SKY, dish
television network which uses MPEG-4,MPEG-1,MPEG-2 and
MPEG-DASH. MPEG-D is being used in audio technologies
like MPEG SURROUND, SAOC (Spatial Audio Object
Coding) and USAC (UNIFIED SPEECH AND AUDIO
CODING)..Speech compression has dynamic adaptive
streaming over HTTP (DASH) technology in the current era of
communication. It is also being widely used in cellular
technology that includes mobile telephony and Voice over IP.
[16]
CONCLUSIONS
Speech compression is a standard for designing and
compressing audio and speech signals which are transmitted to
the receiver end. MPEG technology and its standards define the
compression techniques and provide means for filtering noisy
or undesired signals. Hence, we obtain noise-free speech signal
after decompression of compressed speech signal.
ACKNOWLEDGEMENTS
We are thankful to our college and guides Mrs.Poonam Pathak
and Miss.Vartika Singh respectively for providing time to time
help in studying the topic and providing us a background to
understand it in deeper details

__________________________________________________________________________________________
REFERENCES
[1] Sinha D.,Tewhik A.H, ”Low Bit Rate Transparent
Audio Compression using Adapted Wavelets”, IEEE
TRANSACTIONS ON SIGNAL PROCESSING.
VOL. 41. NO. 12. DECEMBER 1993
[2] Salomon D, "Video Compression: Data compression”
Springer ISBN 978-1-84628-602-5,2007
[3] http://mpeg.chiariglione.org/
[4] ISO/IEC TR 23000-1:2007 – Information technology
– Multimedia application format (MPEG-A) – Part 1:
Purpose for multimedia application formats
[5] ISO. "ISO/IEC 23002-1:2006 – Information
technology – MPEG video technologies – Part 1:
Accuracy requirements for implementation of integer-
output 8x8 inverse discrete cosine transform
technology – MPEG audio technologies – Part 1:
MPEG Surround
technology – Multimedia Middleware – Part 1:
Architecture
technology – Supplemental media technologies – Part
1: Media streaming application format protocols
[9] http://mpeg.chiariglione.org/working_documents.htm#
MPEG-V
[10] Satellite Communications,Seventh Edition
2009,Agarwal D.C,Khana Publications,p-p 368-369
[11] Davis Yan Pan,”Digital Audio Compression”, Digital
Technical Journal Vol. 5 No. 2, Spring 1993
[12] K. Brandenburg and J. D. Johnston, "Second
Generation Perceptual Audio Coding: The Hybrid
Coder," Preprint 2937, 88th Audio Engineering
Society Convention, Montreaux (1990).
[13] K. Brandenburg, J. Herre, J. D. Johnston, Y. Mahieux,
and E. Schroeder, "ASPEC: Adaptive Spectral
Perceptual Entropy Coding of High Quality Music
Signals," Preprint 3011, 90th Audio Engineering
Society Convention, Paris (1991)
[14] Pan, D “A tutorial on MPEG/audio compression”
MultiMedia, IEEE (Volume:2 , Issue: 2 ),2002
[15] Gersho, A. “Advances in speech and audio
compression” Proceedings of the IEEE Volume:82
, Issue: 6 ,August 2002
[16] M. Arjona Ramírez,M. Minami, “Technology and
standards for low-bit-rate vocoding methods,” in The
Handbook of Computer Networks, H. Bidgoli, Ed.,
New York: Wiley, 2008, vol. 2, pp. 447–467
[17] Joshi Anil Kumar,mehta Ashish, “Audio
Compression using Logarithmic Approach for PSNR
enhancement”
BIOGRAPHIES
Mr. Manas Arora, is a M.TECH 2nd
year
student of BBDNIIT college and completed my
B.TECH from SRMCEM,LKO with 77.4%
(HONOURS ) in the year 2012.
Ms. Neha Maurya is a M.TECH 2nd
year
student of BBDNIIT college and completed
B.TECH from SRMCEM,LKO with 74.16%
in the year 2012.
Mrs. Poonam Pathak,is presently Head of
Department of Electronics & Communication
Department at BBDNIIT.She has done her
MTECH in DIGITAL ELECTRONICS &
SYSTEMS.
Ms Vartika Singh is presently working in
BBDNIIT as a senior lecturer. She has done
her Bachelors and Masters Degree in
Electronics and Communication Engineering
from Amity University, Uttar Pradesh.
Presently, she has focused her working area within the various
aspects of the image processing using some suitable simulation
software in the vicinity of implementation of some highly
efficient and accurate algorithms for digital watermarking.Her
research interests also include cooperative communications,
free-space optical communications and wireless
communications

Speech compression analysis using matlab

More Related Content

What's hot (18)

Similar to Speech compression analysis using matlab (20)

More from eSAT Journals (20)

Recently uploaded (20)

Speech compression analysis using matlab