语音谱参数的切换双预测多级矢量量化的码本设计方法

提出了语音谱参数的切换双预测多级矢量量化算法（DPMSVQ） 的码本设计方法。这种改进的多级矢量量化方法充分利用语音谱参数的短时相关和长时相关特性，采用了有记忆的多级矢量量化算法（MSVQ）；并且通过利用相邻语音帧间语音谱参数的强相关和弱相关的不同特点，采用了分别对应于强相关和弱相关的两个预测值，进一步减小了语音谱参数编码位率。切换双预测多级矢量量化方法能够实现21位的语音谱参数近似“透明”量化，同时能够使语音谱参数量化时的计算复杂度略有减少，所需的存储空间大为减少。

Design Method for Switched Dual Predictive Multi-Stage VQ of LPC Parameters

A codebook design method for a switched dual predictive multi stage vector quantization (DPMSVQ) scheme of LPC parameters. This modified multi stage vector quantization algorithm (MSVQ) uses multi stage vector quantization with memory for short term correlation and long term correlation of LPC parameters, and two predictive coefficients for strong correlation and weak correlation of LPC parameters of adjacent speech frames respectively, and reduces the bit rate of LPC parameters. DPMSVQ can achieve transparent quantization of LPC parameters with 21 bits per speech frame, and makes the search complexity and the storage of the quantization of LPC parameters reduced.