摘要
４ｋｂ／ｓ 有限状态代数码激励线性预测语音编码算法ＦＳＡＣＥＬＰ是一种具有延时较短、合成语音质量高、算法复杂度较低的语音编码算法．在线性预测（ＬＰ）参数量化上，利用了语音帧内和帧间的相关性，对线谱对（ＬＳＰ） 参数使用预测式分裂式矢量量化，获得很高的量化效率．在自适应码本搜索上，采用了有限状态控制分数延时搜索的算法，在保证合成语音质量的同时，有效地降低了运算量．对于随机码本，采用了具有多模结构的代数码本，提高语音合成质量．对于激励码序列的增益，采用了预测式矢量量化，有效地提高了量化精度．经非正式听音测试，４ｋｂ／ｓ ＦＳＡＣＥＬＰ的合成语音质量超过了北美８ｋｂ／ｓ ＶＳＥＬＰ，接近Ｇ．７２９ ８ｋｂ／ｓ ＣＳＡＣＥＬＰ，ＭＯＳ分约为３－９ ．
kb/s Finite State Algebraic Code Excited Linear Prediction (FS ACELP) is a high quality speech coding algorithm which has shorter time delay and low calculation burden.For line spectrum pairs,this algorithm uses a highly efficient split vector quantizer based on intra frame and inter frame correlations.For adaptive codebook searching,it uses a finite state fractional delay adaptive codebook which has low calculation complexity and high performence.This algorithm also uses a state controled multimode algebraic codebook and a predictive vector quantizer for gains to further improve the quality of synthesized speech.The non formal tests show that the speech quality of 4kb/s FS ACELP is better than that of IS 54 8kb/s VSELP and close to that of G.729 8kb/s CS ACELP,while the MOS score is about 3 9.
出处
《电子学报》
EI
CAS
CSCD
北大核心
1999年第10期22-26,共5页
Acta Electronica Sinica
关键词
有限状态
多模代数码本
语音编码
FS-ACELP
finite state(FS)
multimode algebraic codebook
vector quantization(VQ)
speech coding
