LCMV分频的改进维纳滤波后置波束形成算法

Improved Wiener post-filter beamforming algorithm based on LCMV divided-frequency

针对封闭环境中语音信号受到混响影响,提出了LCMV分频的改进维纳滤波后置波束形成算法。该算法通过计算麦克风阵列接收到含混响信号的短时傅里叶变换得到频域阵列信号,对频域阵列信号分频处理,将分频的信号进行线性约束最小方差波束形成滤波处理,该波束滤波根据每个频段上混响时间不同的特性对频域阵列信号进行分频处理后,将波束形成算法分别应用到高低频中,以提高混响抑制的精度;再由频域阵列信号的组合功率谱进行维纳后置滤波以抑制混响,由麦克风阵列接收到混响信号的直达波和反射波之间不相关性及麦克风阵列接收信号的空间信息解决维纳滤波器的精确估计问题;最后由逆短时傅里叶变换恢复出时域信号。仿真结果表明,该算法对混响抑制具有明显的改善;且在混响时间600 ms条件下语音增强系统的PESQ值提高了0.26。

In order to solve the problem that the speech signal is affected by reverberation in the closed environment, the improved Wiener filter based on LCMV is proposed. In this proposed algorithm, the frequency domain array signals can be obtained by the Fourier transform of the signals with the reverberation received by the microphone array and are processed according to frequency division method. The divided-frequency signals also are processed by using linear constrained minimum variance beamforming filter method and according to different frequency reverberation time of each beam, and the beamforming algorithm is applied to the high and low frequency domain to improve the accuracy of reverberation suppression for the frequency domain array signals. The combination power spectrum estimation of the frequency domain array signal is used to treat Wiener post-filter in order to suppress the reverberation. Since the received-reverberation signals of microphone array between the direct wave and reflected wave are not related, so we use space information of the received-reverberation signals of microphone array to solve accurately the estimation problem of the Wiener filter. Finally, the time domain signal is obtained by the inverse Fourier transform. The test results show that the proposed algorithm has a significant improvement on the reverberation suppression, and the PESQ score of the speech enhancement system is improved by 0.26 under the condition of reverberation time 600 ms.