摘要
为降低编码速率的同时仍能提供较好的谱失真性能,提出了一种预测分类分裂矢量量化算法,它根据线谱对的特点,融合了预测、分类、分裂的方法对线谱对进行量化,加入了记忆性。实验证明与其他几种方法相比,该算法的量化性能在速率与失真间达到了较好的平衡,且计算量大大降低,仅占有内存有所增加。
In order to reduce bit-rate and still maintain fine distortion performance, this paper proposed predictive switched Split vector quantization method. It added memory according the characteristic of the LSP. And it was a hybrid of switched split vector quantization techniques. Experimental results show that the PSSVQ provides a better trade-off between bit-rate and distortion performance than the others. In addition, the PSSVQ has a lower computational (search) complexity at the expense of an increase in memory requirements.
出处
《计算机应用研究》
CSCD
北大核心
2009年第10期3700-3702,共3页
Application Research of Computers
基金
国家"863"计划资助项目(2007AA012431)
关键词
语音编码
线谱对
多级矢量量化
预测分类分裂矢量量化
谱失真
speech coding
line spectral pair (LSP)
multi-stage vector quantization (MSVQ)
predictive switched splitvector quantization
spectral distortion