Abstract: In AMR-WB, the fixed codebook search is the crucial part in both performance and complexity, and it takes up about 40% of total complexity. To reduce the computational load, this paper proposes an efficient codebook search algorithm based on sub-codeword pulse replacement. The proposed algorithm consists of four steps. Firstly, an initial codeword is divided into two or more sub-codewords. Secondly, each sub-codeword is updated using the least important pulse replacement method. Thirdly, the updated sub-codewords are combined as a candidate codeword. Finally, by comparing the candidate codeword and the initial codeword, better one is selected as the final codeword. Experimental results show that the proposed algorithm reduces the overall complexity by about 16% with minor degradation in speech quality, when compared to the traditional algorithms.

Key words: speech coding, fixed codebook search, codeword split, adaptive multi-rate wideband

摘要： 在AMR-WB中，固定码本搜索是影响性能和复杂度的关键模块，约占总复杂度的40%。为了降低计算量，提出了一种码字分裂、子码字脉冲取代的高效码本搜索算法。该算法包括四步：一个初始码字分裂为两个或更多的子码字；每个子码字通过最不重要脉冲取代法进行更新；更新后的子码字合成一个候选的码字；比较初始码字和候选码字，选择优者作为最后的码字。实验表明，与传统方法相比，编码时间减小约16%。

关键词: 语音编码, 固定码本搜索, 码字分裂, 宽带自适应多码率