CAJ | 학술논문

提出了AMR与G.729语音编码标准之间的2种新型转码算法--索引域转码算法和语音域转码算法.它们分别针对具有相同和不同固定码本结构的语音编码标准进行转码.索引域转码算法直接对2个编码的索引值进行相互转换;语音域转码算法则需要在语音域重新对转换的固定码本及增益进行搜索.实验结果表明,这2种转码算法都能有效地降低转码复杂度,语音域转码算法的算法复杂度仅为传统Tandem转码的55%左右,而索引域转码算法的算法复杂度则不到Tandem转码的10%.同时,索引域转码算法的语音质量相对Tandem转码有所提高,而语音域转码算法则保持了约略相当的语音质量.
제출료AMR여G.729어음편마표준지간적2충신형전마산법--색인역전마산법화어음역전마산법.타문분별침대구유상동화불동고정마본결구적어음편마표준진행전마.색인역전마산법직접대2개편마적색인치진행상호전환;어음역전마산법칙수요재어음역중신대전환적고정마본급증익진행수색.실험결과표명,저2충전마산법도능유효지강저전마복잡도,어음역전마산법적산법복잡도부위전통Tandem전마적55%좌우,이색인역전마산법적산법복잡도칙불도Tandem전마적10%.동시,색인역전마산법적어음질량상대Tandem전마유소제고,이어음역전마산법칙보지료약략상당적어음질량.
Two transcoding algorithms between adaptive multi-rate(AMR) and G. 729 standards, transcoding in index domain and transcoding in speech domain, are presented. They are to transcode between standards with the same and the different fixed codebook structure, respectively. The former translates the indexes between two standards directly, while the latter needs to search the transcoded fixed codebooks and the corresponding gains. The experimental results show that both presented transcoding algorithms can remarkably reduce the transcoding complexity. The computational complexity of transcoding in speech domain is about 55% that of Tandem transcoding, and the computational complexity of transcoding in index domain is even less than 10%. Besides, compared with Tandem transcoding, the quality of transcoded speech is improved through transcoding in index domain, and is maintained through transcoding in speech domain.