利用语音残留冗余的LDPC译码算法  

LDPC decoding algorithm using encoded speech redundancy

作　　者：田秋玲[1] 王童[1] 崔慧娟[1] 唐昆[1] 

机构地区：[1]清华大学电子工程系微波与数字通信技术国家重点实验室,北京100084

出　　处：《清华大学学报（自然科学版）》2006年第7期1254-1258,共5页Journal of Tsinghua University(Science and Technology)

基　　金：国家自然科学基金资助项目(69972020;60572081)

摘　　要：为实现压缩语音在有多径衰落和多用户干扰的无线信道中的可靠传输，提出利用语音码流残留冗余的低密度奇偶校验码（LDPC）译码算法。因信源统计规律的时变性及信源编码复杂度和延时的限制，语音编码输出码流中还存在残留冗余，在信道译码时利用这些冗余可有效增强LDPC迭代译码的纠错能力。仿真结果表明：通过利用语音码流中的残留冗余，能有效提高LDPC码的纠错性能，减少平均迭代次数，明显降低译码运算量，改善合成语音质量。特别在信噪比较低时，效果更加显著，平均迭代译码次数下降30％～50％，合成语音平均谱失真下降0．3～0．8dB。A joint source-channel decoding （JSCD） scheme based on low-density parity-check code （LDPC） was developed for reliable transmission of compressed voice through a wireless channel with muhi-path fading and multi-user interference. The LDPC decoding algorithm exploits the residual redundancy inherent in the encoded speech stream. Ideal compression can not be realized due to source nonstationarity and complexity restrictions and source coding delays. Simulation results show that, by exploiting the residual redundancy in the code stream after speech coding in the LDPC decoder, the scheme can effectively improve the error-correction performance, greatly reduce the average number of iterations and computational complexity and enhance the quality of the reconstructed speech. The improvements are very significant with medium and low signal-to-noise ratio, where the average number of iterations is reduced by 30% to 50% and the average spectrum distortion of the synthesized speech is reduced by 0. 3 to 0. 8 dB.

关 键 词：低密度奇偶校验码(LDPC) 信源信道联合译码 和积译码算法 残留冗余 语音编码 

分 类 号：TN912.3[电子电信—通信与信息系统]