论文部分内容阅读
在低速率语音编码算法中,如何对特征参数进行有效的量化表示是影响声码器合成语音质量的关键因素。该文提出一种能量参数解码端恢复算法,它利用线谱频率(linespectral frequency,LSF)和清浊音判决参数(unvoiced/voiced decision,U/V)估计能量参数的变化轨迹。该算法利用特征参数之间的相关性,采用隐Markov模型(hiddenMarkov model,HMM)描述LSF、U/V和能量参数之间的统计特性,通过对能量进行解码端恢复,省去量化所需的比特数,从而提高特征参数的整体量化性能。测试结果表明:能量参数解码端恢复算法能够将150b/s混合激励线性预测编码算法(mixed excitation linear prediction,MELP)的合成语音平均意见得分(mean opinion score,MOS)提高0.042。该算法应用于超低速率声码器参数量化是可行的。
In the low-rate voice coding algorithm, how to effectively quantify the characteristic parameters is a key factor that affects the vocoder’s voice quality. In this paper, an energy parameter decoding end recovery algorithm is proposed, which uses the linespectral frequency (LSF) and unvoiced / voiced decision (U / V) to estimate the trajectory of energy parameters. The algorithm uses the hidden Markov model (HMM) to describe the statistical properties of LSF, U / V and energy parameters by using the correlation between the characteristic parameters. By decoding the energy side, the algorithm eliminates the need for The number of bits increases the overall quantization performance of the feature parameters. The test results show that the energy parameter decoder recovery algorithm can improve the mean speech score (MOS) of synthesized speech of 150b / s hybrid excitation linear prediction (MELP) by 0.042. It is feasible to apply this algorithm to ultra-low-rate vocoder parameter quantization.