Complexity Reduction for IETF-iLBC Speech Coder

• Main Authors: Kuo-Bao Kuo, 郭國寶
• Other Authors: Fu-Kun Chen
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/83293637830643107097

碩士 === 南台科技大學 === 資訊工程系 === 97 === With the development of the Internet, the needs to transmit information through the Internet have also increased. To find a low-rate but high-quality speech coding technology is the major issue of the speech communications over Internet. The iLBC speech codec has been recommended by the international organization, Internet Engineering Task Force (IETF). In the iLBC encoder, the dynamic codebooks, included the base-codebook, the augmented base-codebook, the expanded codebook and the augmented expanded codebook, bear the maximum loads of coding procedure. Especially, the computational loads of the base-codebook and the expanded codebook account for the largest volume of computational complexity in the codebook search. This thesis aims to reduce the computational complexity with perceptually negligible degradation of the speech quality. The pattern-pre-inspecting search method is therefore proposed. The proposed approach can indeed reduce the computation of the codebook search. In addition, this thesis proposed a complexity scalability design for adaptive codebook search for the hand-held or mobile devices so that the coding process can be adjusted dynamically according to the computation loading. According to the evaluation of the MOS-LQO score, an objective measure of ITU standards, the proposed approaches not only substantially reduce the computational complexity, but also maintain good quality of the coded speech. Furthermore, because the writing format of IETF differs from the international standards, it is difficult for engineers to implement and read. The thesis also illustrates the iLBC coding process in detail, and rewrites the process in the ways of ITU and MPEG standards.