关于多通道语音去噪的识别优化研究

Recognition and Optimization of Denoising Method for Multi-Channel Speech

在语音信号优化识别中,由于多通道语音受到噪声的污染,效果较差。为提取纯净的语音信号,通常采用传统的小波方法。虽然小波方法具有很强的去噪能力,但是会在语音信号去噪的同时造成有用信息的丢失,从而导致去噪的效果不理想。为了解决上述难题,提出了一种多通道语音去噪系统,对语音信号进行去噪。首先利用麦克风阵列、时延估计、时延补偿、加权求和来对多路语音信号进行采集和预处理;然后再通过小波对预处理后的语音信号进行分解,提取出部分噪声作为RLS自适应滤波的参考输入。经过RLS自适应滤波处理后,使得RLS自适应滤波输出信号和语音信号中的噪声部分具有很好的相关性,最后对噪声进行抵消处理。仿真结果表明,文中改进方法比传统的小波去噪效果更好,可为语音信号优化识别提供科学参考。

Due to the poor quality of the multi - channel speech by the noise pollution, traditional wavelet method is easy to cause the useful information lost in the speech signal denoising. In order to solve the problem, a denoising method to multi - channel speech based on recognition and optimization is presented. First of all, the multi - channel speech signal is acquired and preprocessed by using microphone array, time delay estimation, time delay compensa- tion, and weighted sum. Then the noise is extracted from the speech signal processed by wavelet. The extracted noise is used as the reference input of the RLS adaptive filter. The noise in the output signal of the adaptive filter and the noise in the speech signal have a good correlation after the RLS adaptive filter processing, and the noise is offset by the processing. The simulation results show that the improved method is more effective than the traditional wavelet denoising method.