软判决修正下的语音增强算法在数字信号处理器上的实现与优化

摘要

针对低信噪比情况下,传统谱估计语音增强算法性能下降的问题,提出了一种结合语音存在概率和人耳听觉掩蔽效应的软判决短时谱估计算法.该算法在低信噪比的情况下保护微弱的语音信号,减少语音谱的失真.同时介绍了基于TMS320C5502数字信号处理器的嵌入式系统,重点讨论了系统总体方案和主要功能模块的硬件设计,并且结合数字信号处理器的结构特点,给出了该算法系统平台移植和优化的方法.系统测试表明,硬件平台工作稳定可靠,针对硬件资源的移植和优化显著地提高了系统的处理速度,满足实时性的要求,且输出语音在去噪度与失真之间很好地平衡,优于传统的谱减算法.%The authors proposed a short-time spectral amplitude estimation approach based on speech absence probability information and the masking properties of the human auditory system to improve the performance of speech enhancement algorithm in low signal to noise ratio input environments. Meanwhile, the hardware design and the algorithm optimization based on the TMS320C5502 digital signal processor embedded system were introduced. The system testing results show that the hardware platform works stably and reliably, and the optimization can significantly improve system processing speed. Besides,the output signal achieves a good balance between noise reduction and speech distortion.