【摘要】 对转换分类分裂矢量量化算法进行了深入研究。通过调整量化算法中矢量的分裂方式及量化所需比特数的分配,实现出了一种应用于宽带ISF参数量化的改进的42 bits/frame转换分类分裂矢量量化方案。该方案采用无记忆的量化方式,并根据ISF参数的特性提出了一种新的加权失真测度计算方法。实验表明,该方案在每帧编码比特数为42时达到了透明量化的性能,且在发生帧丢失的情况下不会产生错误繁衍。更多还原

【Abstract】 Some researches focused on switched split vector quantization(SSVQ) are shown.By adjusting the split form and the bit allocation,an improved 42 bits/frame switched split vector quantizer for quantizing immittance spectral frequency(ISF) parameters of wideband speech is realized.The SSVQ is a memoryless quantization scheme.And a new weighted distance measure on the characteristic of the ISF parameters is proposed.Experimental results show that the quantization scheme achieves transparent coding at 42 bits/frame,and there is no error propagatation in frame erasure.更多还原