论文部分内容阅读
声码器的增益参数表示语音的能量。误码导致的增益错误,会造成合成语音的响度失调,产生刺耳毛刺和过载等问题,严重影响听觉质量,甚至损伤听力。为提高其鲁棒性,该文针对增益参数提出了一种基于子带清浊音模式的解码端抗误码算法,利用增益参数在不同清浊音模式下的分布统计特性和不同清浊音转移模式下的变化差值特性,在解码端对信道误码引起的增益参数错误进行检测和恢复。算法也考虑了错误扩散的防止和对不同响度水平语音的自适应等问题。实验结果表明,在2.4 kb/s和1.2 kb/s速率的正弦激励线性预测(SELP)声码器上,该算法可以在无额外比特开销的条件下,将增益参数由于信道误码而引起的错误方差降低30%以上。
The vocoder gain parameter represents the energy of the speech. Errors caused by bit errors can cause loudness imbalance in the synthesized speech, resulting in glitches and overload, which can seriously affect hearing quality and even impair hearing. In order to improve its robustness, this paper proposes a decoding end-error resilience algorithm based on sub-band unvoiced and voiced mode for gain parameters. By using the statistical properties of gain parameters under different clear and voiced modes and different clear and voiced transition modes The difference between the characteristics of the changes in the decoding side of the channel error caused by error parameter detection and recovery. The algorithm also considers such issues as the prevention of false diffusion and the adaptation of speech at different loudness levels. The experimental results show that the proposed algorithm can compensate gain parameters due to channel error without additional bit overhead at 2.4 kb / s and 1.2 kb / s sinusoidal excitation linear prediction (SELP) vocoder The error variance is reduced by more than 30%.