Title :
A fast search technique for multistage vector quantization based on the introduction of tree-structure to each stage
Author_Institution :
Media Lab., DoCoMo USA Labs., San Jose, CA, USA
Abstract :
In this paper we describe a method to reduce the search complexity required to locate the codevectors during encoding in multistage vector quantization (MSVQ). The core idea is to add a series of codebooks so as to support a tree-structure at each stage. By searching with the tree-structure, substantial complexity reduction can be achieved. The codebook design algorithm, search technique and experimental results are provided, which are compared to traditional techniques to show the advantages of the proposed method.
Keywords :
speech coding; trees (mathematics); vector quantisation; MSVQ; codebook design algorithm; encoding; multistage vector quantization; search technique; speech coding; tree-structure; Acceleration; Algorithm design and analysis; Design optimization; Encoding; Laboratories; Tree data structures; USA Councils; Vector quantization;
Conference_Titel :
Signals, Systems and Computers, 2004. Conference Record of the Thirty-Eighth Asilomar Conference on
Print_ISBN :
0-7803-8622-1
DOI :
10.1109/ACSSC.2004.1399167