基于共轭结构ACELP语音编码算法的优化及DSP实时实现

使用高质量、高压缩率的语音编解码器,可以在保证通话质量的前提下有效地提高通信系统的容量。这里首先介绍了基于共轭结构ACELP语音编解码算法,并针对该算法特点提出算法优化方法,如固定码本搜索优化,开环基音分析优化,运算移位溢出检查处理等;然后给出在TI公司16位定点DSP芯片TMS320VC549上实时实现该优化后算法的方案和过程,分析了编解码语音的质量和算法的性能,最后得出实时实现的结果。实验结果表明,优化后的CS-ACELP语音编码算法在定点DSP上的实时实现,能得到良好的语音音质,所用的资源和功耗极低。

REAL-TIME IMPLEMENTATION OF OPTIMIZED CS-ACELP SPEECH CODEC ON FIXED-POINT DSP

In this paper,the ACELP speech coding algorithm which is based on Conjugate-Structure(CS-ACELP) is briefly described and some optimization methods are proposed,and the scheme and procedure of the algorithm's real-time implementation on the TI TMS320VC549 fixed-point DSP chip are also presented.Then,the quality of the original speech and reconstructed speech,and the performance of the speech coder are analyzed and some subjective test results are also presented. The measurement results show that the real-time implementation of optimized CSACELP speech coder on fixed-point DSP can produce toll-quality speech with smaller amount of resource and lower power cost.