论文部分内容阅读
针对低速率语音编码领域的正弦激励线性预测(SELP)声码器,提出了一种错误帧参数修复算法。对于信道编解码不能纠错的语音帧,算法根据信道解码后所指示的错误信息,结合解码端各参数平稳性的不同,采用参数替换方法分别进行修复。测试结果表明:采用该算法,在1%、3%、5%的随机信道误码下,声码器抗误码性能均有较大改善,特别是在5%的较高信道误码率下,合成语音平均意见得分提高了0.2以上。
For sinusoidal excitation linear prediction (SELP) vocoder in the field of low-rate speech coding, an error frame parameter restoration algorithm is proposed. For a speech frame that can not be corrected by channel coding and decoding, the algorithm repairs by using the parameter replacement method according to the error information indicated after channel decoding and the stability of each parameter at the decoding end. The test results show that the proposed algorithm can greatly improve the error resilience performance of the vocoder at random channel errors of 1%, 3% and 5%, especially at a higher channel error rate of 5% , The average opinion score of synthesized speech increased by more than 0.2.