一种基于维纳滤波去除语音通信中混响的方法

A Means Based on Wiener Filtering for Dereverberation in Speech Communication

简要叙述了语音通信中去混响的意义.利用常规通话起始语音的习惯特点,提出了预存起始纯净语音信号,并基于维纳滤波原理,通过反卷积运算求出房间冲击响应,再通过反卷积去除语音信号混响的新方法.对语音样本的仿真试验表明,该方法对单字语音的去混响效果良好,在普通计算机上每字的去混响运算耗时为0.3-0.5 s.

This paper mentioned briefly the significance of dereverberation in speech communication. The habitual greeting words of usual talk are utilized in obtaining room impulse response function. Other words in the talk are then de-reverberated by de-convolution with the obtained room response function through Wiener filtering. Implementation of the means word by word to speech samples from speech library demonstrates a satisfying dereverberation effect, with the average time cost about 0. 3 to 0. 5 s per word on ordinary computer.