266
DIFFERENT SPEECH COMPRESSION TYPES IN
SIMULTANEOUS INTERPRETING
Ibragimova Dilnora Shavkat qizi
Master student, Uzbekistan State World Languages University
This publication paper describes the different types of speech compression
techniques. First of all we are eager to give the definition of the notion “speech
compression”. The aim of speech compression is to reduce the number of bits
required to represent speech signals by removing the redundant bits so-that the less
bandwidth is required for transmission. Before discussing the speech compression
coding techniques, it is important to understand the digitization process. The speech
signal is represented in its digital form, that is, the process of speech signal
digitization. There are basically two the key features of speech signal are voiced and
unvoiced speech and their characteristics. In broader terms, speech compression
techniques are mainly focused on removing short-term correlation (in the order of
1ms) among speech samples and long-term correlation (in the order of 5 to 10 ms)
among repeated pitch patterns. In this section, we will start with speech signal
digitization and then discuss speech signal features and speech compression
techniques.
Speech compression is use in the encoding system. The bit rate reduction is
use in the encoding system. By the use of bit rate reduction algorithm, the minimum
bits are used to compare the original information.
There are different speech compression techniques are present. Basically is
divided in two types:
Lossy
Lossless.
Lossy compression means a class of data compression algorithm that allows
the exact original data to be reconstructed from the exact original data to be
reconstructed from the compressed data but bit rate and is better than lossless. It is
compression ratio is higher than lossless compression. While lossless means output
signal and input signals sounds undistinguished. Speech coder analysed using
subjective and objective analysis. Subjective is making judgments by listening output
and original signal. Playing back signal and checking quality. Objective includes
technical assess. Such as computing segmental signal to noise ratio (SEGSNR)
between original and output signal. Speech compression motivation is to remove
redundancy in speech representation to reduce transmission bandwidth and storage
space or memory (and apart to reduce cost). The purpose of speech compression is
to reduce the number of bits required to rep- resent Speech signals (by reducing
redundancy) in order to minimize the requirement for transmission bandwidth (e.g.,
for voice transmission over mobile channels with limited capacity) or to reduce the
storage costs (e.g., for speech recording). Before we start describing speech
compression coding techniques, it is important to understand how speech signal is
represented in its digital form, that is, the process of speech signal digitization. There
267
are in general three basic speech compression techniques, which are waveform-
based, parametric based and hybrid coding techniques.
WAVEFORM BASED SPEECH CODING
Waveform-based codecs are intended to remove waveform correlation
between speech samples to achieve speech compression. It aims to minimize the
error between the re- constructed and the original speech waveforms. It is classified
as time domain and frequency domain
Time domain: such as A. PCM (Pulse code modulation) B. ADPCM (Adaptive
Differential PCM)
Frequency domain or Transform coding: such as A. Fast Fourier Transform
(FFT) B. Discrete Cosine Transform (DCT) C. Continuous Wavelet Transform (CWT)
D. Discrete Wavelet Transform (DWT) Waveform coders are able to produce original
signal at decoder (Lossless). Bit rate range
– 64 kb/s to 16 kb/s. At bit rate lower than
16 kb/s, the quantization error for waveform based speech compression coding is too
high, and this results in lower speech quality.
PARAMETRIC-BASED SPEECH CODING
Parametric-based compression methods are based on how speech is
produced. Instead of transmitting speech waveform samples, parametric
compression only sends relevant parameters related with speech production to the
receiver side and reconstructs the speech from the speech production model. Thus,
high compression ratio can be achieved. Bit rate range
– 1.2 kb/s to 4.8kb/s
Many different techniques are explored to represent waveform-based
excitation signals such as multi-pulse excitation, codebook excitation and vector
quantization. The most well known one, so called Codebook Excitation Linear
Prediction (CELP)‖ has created a huge success for hybrid speech codec in the range
of 4.8 kb/s to 16 kb/s for mobile/wireless/satellite communications. Types of Hybrid
speech compression: A. Codebook Excitation Linear Prediction (CELP) B. Vector
Sum Excited Linear Predictive Coder (VSELP)
Today, many compression techniques are developed and some techniques
are in process. But this paper only discusses the general idea about the Waveform-
based speech compression, Parametric-based speech compression and Hybrid
based speech compression. Parametric based codec is higher in implementation
complexity but can achieve better compression ratio. This paper has been written to
understand Parametric-based speech compression in the better manner and relate
it to the future work.
As the data grows day by day, the short and compression Communication is
required, Speech compression techniques plays a vital role in it. In this paper all the
possible speech compression techniques are discussed that can be used to
compress the data before transmission of data so that it can consume less
bandwidth.
REFERENCES:
1.
EmrahAkyol, Kenneth Rose, “A Necessary and Sufficient Condition for
Transform Op
timality in Source Coding”, IEEE Int. Symp. on Information Theory
Proceedings, 2011.
– PP. 2597–2601.
268
2.
J.N. Holmes “The JSRU Channel Vocoder” IEEE Proc., Vol: 127 – PP: 53–60,
1980. 3.J. Lukasaik, C.
McElory, E. Chang, “Compression Transparent Low-Level
Description of Audio Signals”, IEEE Int. Conf. on Multimedia and Expo, 6 July 2005.
4. Jerry D. Gibso
n, “Speech Coding Methods, Standards, and Applications”,
Department of Electrical & Computer Engineering University of California, Santa
Barbara, CA 93106
–6065.
5. Saumya Mishra, Shraddha Singh “A Survey Paper on Different Data
Compression Techniques”, Indian journal Research Paper, Volume: 6, Issue: 5,
May 2016.
6.
RajeshreeRaut& Kishore Kullat, “SDR Design with Advanced Algorithms for
Cognitive Radio” IJACS, Vol.1, No.4, pp: 134-141, Oct 2011.
7.
Tejaswi Nanjundaswamy, Kenneth Rose, “Cascaded Long Term Prediction
for Coding Polyphonic Audio Signals”, IEEE w/s on Applications of Signal Processing
to Audio and Acoustics, New Paltz, NY, Oct. 2011.