【摘要】 语音信号去混响技术能明显提高语音通信系统和识别系统的性能。简要介绍了语音复倒谱法去混响原理,对复倒谱域去混响滤波特性进行了仿真研究。根据多种去混响评价指标,确定复倒谱域"低通滤波器"的最高截止点、过渡带宽和过渡带的曲线特性等参数,发现在通常混响时间范围内,"低通滤波器"最高截止点与混响时间无关,复倒谱域滤波前加高斯窗可以改善去混响效果。更多还原

【Abstract】 Speech signal dereverberation may greatly improve the performance of speech communication system or recognition system.The principle of dereverberation by cepstral processing is briefly introduced,and numerical simulation research is conducted to investigate filter properties in cepstral domain.By several dereverberation evaluation criterions,the cut-off points of the lowpass filter in complex cepstral domain are determined,and the transition bandwidth and transition section’s curve characteristic are also fixed.It is shown that,the cut-off point of the lowpass filter is independent of the reverberation time in common reverberation time.It is also found that the dereverberation effect will improve if a Gauss window is used before complex cepstral filtering in time domain.更多还原