一种基于TMS320C6713的语音信号频域盲分离算法

Blind Speech Separation in Frequency Domain Based on TMS320C6713 DSK

提出了一种真实环境中语音信号频域盲分离算法的实现方案。基于TMS320C6713 DSK平台,充分利用硬件资源,完成了语音盲分离处理算法的平台移植,并进行了编译环境、算法和程序3个级别上的软件优化,最终结果显示本系统具有较好的性能,在离线和在线情况下均取得了对真实环境中两说话人混合语音的有效分离。

An implementation for speech blind source separation of frequency-domain in a real environment is introduced. Based on the TMS320C6713 DSK platform, the hardware resources are used, and the blind speech separation on the DSP platform is implemented. Meanwhile, software optimizations on the levels of compiling environment, algorithm and program are con- ducted. Experimental results show good performance on two-speaker mixed speech separation in a real room on offline and online scenarios.