一种空域、时域与频域相结合的阵列语音增强算法

提出了一种空域、时域与频域相结合的阵列语音增强算法，对含噪信号进行处理：第一步，先在空域及时域内利用波束形成得到一个初步增强的语音信号；第二步，利用前面得到的增强信号，在时域内采用多通道自适应滤波消除相关噪声；第三步，在频域内利用谱相减消除残余噪声。整个算法的消噪量大，而且阵列的孔径小，能应用于多种噪声环境。

An Array Speech Enhancement Algorithm Based on the combination in Spacial, Time and Frequency Domain

A multimicrophone array with the algorithm that combined in spacial domain, time domain and frequency domain is proposed in this paper. The proposed system for noise reduction is processed in three steps: Beamforming is firstly used to get the enhanced speech, then it is applied to the second step, in which an adaptive multichannel filter as the desired component, thus the possible coherent noise components are canceled. The residual additive noise components are reduced in frequency domain by spectral subtraction. The scheme proposed can remove quite a lot of noise with a small size of the array, thus it can be used to many kinds of noise cancellations.