A Novel Vector Quantizer Design Method:the Codebook Reorganization Algorithm

• Main Authors: Yuan-Tsung Chen, 陳源聰
• Other Authors: Ting-Wei Hou
• Format: Others
• Language: zh-TW
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/6s3sq9

碩士 === 國立成功大學 === 工程科學系碩博士班 === 92 === 　　The research proposed a novel approach named CRA(Codebook reorganization Algorithm), for the vector quantizer design. The vector quantizer design involves a clustering process of many training vectors. It is an application of the data clustering problem in natural. The complexity of algorithms for solving the data clustering problem has been proved to be NP-complete. 　　GLA algorithm is the most famous known algorithm in the field of vector quantizer design. It runs very fast ,but it’s ability to find the global optimum is poor. In most cases, it can only find a poor local optimum. In 1989, Zeger proposed the SRD algorithm to overcome the weakness of the GLA algorithm. Theoretically, it has the ability to find the global optimum. In practice, it can achieve the near optimal performance 　　The performance of CRA algorithm is superior to the above two algorithms demonstrated by the experimental results. It can find better codebooks then GLA. It can find a codebook as good as SRD in less time.