论文部分内容阅读
为了在存储量受限的情况下尽可能提高线性预测编码(linear predictive coding,LPC)系数量化性能,提出了一种基于码本共享算法的分模式多级矢量量化(multi-stagevector quantization,MSVQ)算法。由于LPC参数的分布与清浊音(unvoiced/voiced,U/V)参数相关,该算法对不同U/V对应的LPC参数进行不同量化,然后利用码本共享算法减少存储量需求。实验表明:在相同码率的情况下,该算法较MSVQ平均谱失真(spectrum distortion,SD)降低3.2%,码本大小增加26.7%;较分模式量化(mode-basedquantization,MBQ)平均谱失真升高3.6%,但是码本尺寸下降了92.1%。该算法是MSVQ与MBQ算法的一种折衷,在增加少量存储量的情况下提高了LPC系数的量化性能。
In order to improve the quantification performance of linear predictive coding (LPC) coefficients under the condition of limited storage capacity, a novel multi-stage vector quantization (MSVQ) algorithm based on codebook sharing algorithm is proposed. algorithm. Since the distribution of LPC parameters is related to unvoiced / voiced (U / V) parameters, this algorithm quantizes the LPC parameters corresponding to different U / Vs differently and then uses the codebook sharing algorithm to reduce the storage requirement. Experiments show that the proposed algorithm reduces the average spectrum distortion (SD) by 3.2% and the codebook size by 26.7% at the same bit rate. The mean spectral distortion of mode-based quantization (MBQ) 3.6% higher, but the size of the codebook dropped by 92.1%. This algorithm is a compromise between MSVQ and MBQ algorithm, which improves the quantization performance of LPC coefficients with a small amount of memory added.