Yao-Tang Chou, Wen-Liang Hwang and Chung-Lin Huang
Abstract
We show that the techniques of the shape-gain product vector quantizer(VQ) can be used to optimize the dictionary of the matching pursuit codec and produce good results. We start with an initial dictionary and obtain a dictionary that is wellz adapted to the training sequences. The performance is evaluated based on a comparison of our method with other methods.
¡@