Speaker sound coding using vector quantization technique (Vq)‎

Abstract EN

The key objective of this research is to compress the speech sound or in another meaning coding speaker sound with a small matrix called code book represent the speaker sound information this is done by using Vector Quantization method (VQ).The sound features here represented with both types Linear Predictive Coding (LPC) and autocorrelation coefficients that represent the data base of this research.

The algorithm is tested upon a database consist of matrix with dimensions (54 * 13) which represent speaker sound information with normalized autocorrelation coefficients and compress it to a codebook (CB) contain 8 codeword (8 CW) with autocorrelation coefficients and the other CB with LPC coefficients of order p = 12.

From this project we can notes how much the sound information can be compressed which represented at first with a matrix of dimensions (54 * 13) and transfer it to a matrix with dimensions (8 * 12) or (8 * 13) with same feature sound of speaker.

In this project there is no focus on LPC method and how it work but used it in extracting Sound features in another program which represent data base for this project and inter it to VQ algorithm The algorithm was examined through computer simulation using Matlab version 6 programming language and under Microsoft Windows XP operating system.