Khruschak S. Vector quantization of signal parameters for speech compression in computer systems

Українська версія

Thesis for the degree of Candidate of Sciences (CSc)

State registration number

0412U002977

Applicant for

Specialization

  • 05.13.05 - Комп'ютерні системи та компоненти

19-05-2012

Specialized Academic Board

Д 05.052.01

Vinnytsia national technical university

Essay

Object of research - the process of processing, transmitting and storing of speech information in computer systems; the purpose of work is to improve speech signals processing performance in computer systems by developing software and hardware for data compression necessary for further transmission, storing and playback of speech signal while keeping acceptable quality of decoded speech, methods of the digital signal processing theory, methods of the mathematical statistics theory, computer modeling were used, theoretical results - the method of speech signal parameters delta-compression is proposed, which, unlike the existing differs in that it transmits delta of current and predicted vectors indexes in structured code book, which allowed to reduce amounts of data transmitted and stored in a computer system, method of structuring the code book is improved, which, unlike the existing uses arrangement based on the principles of vectors mazorization, which allowed to reduce search time in the process of vectors quantization in a low speed speech signals processing computer system, adaptive voice activity detection method improved, which, unlike the existing uses geometric adaptive threshold and smoothing of the spectral envelope, thus improving the performance and accuracy of computer systems under conditions of low signal to noise ratio. The practical results - algorithms and software for speech signals compression based on vector quantization are developed, algorithms and software tools for fast vectors search methods are developed, the architecture of speech signals processing computer system with proposed methods support is developed. Introduction degree - developed adaptive method of voice activity using smooth spectral envelope and geometric threshold and advices about using methods of audio compression are applied at the Institute of electronics and communications of the Ukrainian Academy of Sciences. Sphere of use - computer systems.

Files

Similar theses