基于泄漏约束的改进型语音增强算法

An improved speech enhancement algorithm based on leakage constraint

针对广义旁瓣抵消器算法(GSC)对声源定位误差敏感的缺点,提出了一种改进的语音增强方法:改变GSC结构中的自适应算法,避开声源定位的问题,避免了因信号来向角失配带来的信号泄漏,使算法具有较好的顽健性;将维纳滤波结合到GSC的非自适应支路中,有效地消除了非相干噪声.仿真结果表明,改进后的GSC算法比经典的GSC波束形成算法和带反馈的GSC算法具有更好的噪声抑制性能,并且在语音可懂度方面也有明显改善.

Aiming at conventional generalized sidelobe canceller（GSC）＇s sensitivity to source positioning error,an improved algorithm based on generalized sidelobe canceller was proposed.By avoiding the source position and changing the structure of the GSC adaptive algorithm,it can avoid signal cancellation,which makes the improved algorithms have better robustness.In addition,combining a wiener filter with the non-adaptive GSC structure will eliminate incoherent noise effectively.The simulation results showed that the improved GSC algorithm exhibited better noise suppression performance compared with conventional GSC algorithm and decision feedback-GSC algorithm and better speech intelligibility.