readme.txt

VQ压缩算法

THREE  VQ CODEBOOK TRAINING ALGORITHMS

Copyright (c)  by Jialong He, all rights reserved.


Introduction:
-------------

In this package, I implemented three codebook training 
algorithms, including the GVQ,  the LBG and the LVQ. The 
optimization criteria of these three algorithms are different. In 
LBG training, the codebook is optimized to give the smallest 
quantization error for the training vectors. This criterion is 
suitable for the applications such as speech coding. One the 
other hand, LVQ trained codebook can give a better 
classification performance for individual feature vectors. 

In speaker identification, a classification decision is made 
based on a number of feature vectors. Therefore, in GVQ 
training, the codebook is optimized to give a lower 
classification error rate for 搗ector groups