基于线谱对高效矢量量化的0.6kb/s语音编码算法

为了提高通信系统的抗干扰和抗攻击能力,尽可能降低语音编码速率.提出了一种O.6 kb/s语音编码算法.算法基于3帧联合,对多帧联合参数采用高效矢量量化,在降低语音编码速率的条件下保证语音编码质量.其中,对线谱对参数采用预测多模式多级矢量量化码本结构.在码本设计过程中,提出了多模式渐进闭环设计,对各类码本联合优化,并联合优化预测器和量化器,可以有效提高线谱对参数量化质量.在译码方,采用多带混和谐波激励提高合成语音清晰度.测试结果表明,该语音编码算法合成语音PESQ(perceptualevaluation of speech quality)得分可以达到2.7,汉语诊断押韵测试DRT(diagnostic rhyme test)得分可以达到89.7.

Speech coding algorithm at 0. 6kb /s based on efficient vector quantization of linear spectrum pains parameters

A speech coding alg ori thm a t 0. 6 kb /s w as presented to satisfy disadv antaged transmissio n requirements. The pa rameters of co nsecutiv e 3 f rames w ere g rouped tog ether and joint ly qua nti zed to improve the speech quali ty a t low bit rate. LSF pa rameter w as quanti zed wi th multimode predictiv e multistag e v ecto r quantization and the codebook designed wi th mul timode asympto tically clo sed-loo p desig n. The nov el aspect o f the propo sed approach w as joint o ptimi zatio n of the multimode codebook co nsidering the mode transitio n betw een successiv e superf rames. At the decoder, the speech wa s synthesized w ith a proposed mul tiba nd mix ed ha rmonic ex citatio n model. The resul ts show that the PESQ score can achiev e 2. 7 point s a nd th e DRT sco re ca n achiev e 89. 7.