基于MGSC与改进维纳滤波的麦克风阵列语音增强

针对传统广义旁瓣抵消中阻塞矩阵语音泄露、非相干噪声消噪能力较弱及后置维纳滤波中相位不变等问题,提出一种基于改进广义旁瓣抵消与相位补偿维纳滤波的方法。该方法将阻塞矩阵变为阻塞滤波器从而减少了阻塞矩阵语音泄露,然后将相位补偿的维纳滤波用于估计纯净语音的幅度谱和相位谱,从而抑制广义旁瓣抵消残留的噪声。仿真及实测结果表明,该方法能够更加有效地抑制噪声的影响,提高语音的可懂度。

Speech enhancement of microphone array based on MGSC and improved Wiener filter

Traditional generalized sidelobe canceling algorithm is bad for non-coherent noise with the issue of speech leakage in block matrix, while Wiener post-filter also has the problem on unchanged phase. In order to modify this, a method is proposed based on modified generalized sidelobe canceling and phase compensation Wiener filtering for speech enhancement of microphone array. This method applies the idea of using a block filter instead of the block matrix to reducing speech leakage in block matrix, and meanwhile a way of Wiener filtering based on phase compensation is used to improve the performance of the post-filtering after generalized sidelobe canceling via estimating the magnitude spectrum and the phase spectrum of clean speech. The simulation results and perceptual evaluation of speech quality show that this method is able to eliminate the noise and to improve the speech intelligibility more effectively.