الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 185 |  |  |
| | | from | 185 | | |
Abstract
?The problem reported here is how to transmit a digital speech signal using different techniques within a small bandwidth in order to use an analog channel. Among the available techniques, adaptive coding has shown considerable potential in providing high quality speech at bit rates below 20 kbps. The encoding and transmission of speech at these low bit rates is of considerable importance in military and mobile applications for secure speech transmission and could also be used for improving the quality of conventional telephone links. A pulse code modulation (PCM) with both linear and adaptive quantisers is adopted in part of the study in this thesis. It is found that using Wavelet and Haar transforms upon the speech signal provide more bandwidth compression than the conventional PCM. ?Another technique adopted in this study is the differential pulse code modulation (DPCM) system for speech coding at low bit rates and high compression ratio. Better performance is obtained using adaptive predictor to remove the redundancy present in the speech signal. Adaptive predictive coding aims to remove redundancy from digitised speech by means of an adaptive digital filter whose frequency response is updated as the shortterm spectral properties of the speech change with time. The adaptive filter output, prediction error, and parameters which specify the frequency response are then coded and transmitted as a low bit rate representation of the input speech. The speech may be reconstructed by passing the prediction error signal through a digital ?synthesis? filter whose transfer function is the inverse of the adaptive prediction filter transfer function. An alternative approach is to use sequential adaptive linear prediction methods based on ?samplebysample? updating of the adaptive prediction filter. It is noted that using the samplebysample prediction filter upon the Wavelet or Haar transformed speech signal provides a very small prediction residual which can be quantised using 1 bit/sample. That is a very large compression ratio with good quality reproduced speech. ??In order to improve the performance of PCM system, vector quantisation has been introduced to encode the speech signal directly. Moreover, Wavelet and Haar transformed speech are also quantised using vector quantisation. It is found that, PCM system with vector quantisation to the Wavelet or Haar transformed speech provides the best SNR as compared to the above mentioned techniques at bit rate of 10 kbps ( compression ratio 91.67%).