传声器阵列波束比判决语音增强方法
A speech enhancement approach based on power level ratio of multi-beamformers
-
摘要: 针对单一波束形成器难以深度抑制空间相干干扰的问题,提出了一种综合了最小方差无畸变响应波束形成器与对称子阵延时求和波束形成器的语音增强方法。定义了一种波束输出比因子,根据该因子在目标声区域和干扰声区域的幅值变化,给出了采样协方差矩阵对角加载量的调整方法,并进一步利用该因子在后滤波环节对空间干扰进行判决滤波。文中对判决滤波时的上限阈值和下限阈值的实时更新方法给出了说明。所提出的算法能进一步抑制空间干扰和噪声,且可满足实时需要。在传声器圆阵上的实验表明,该方法在输出信干噪比及语音质量上,均优于经典对角加载算法及采样协方差矩阵扫描重构算法。Abstract: To improve the performance of the beamformer on spatial interference suppression and noise reduction, a method combining the Minimum Variance Distortionless Response (MVDR) beamformer and the Delay and Sum (DAS) beamformer is introduced. Based on the output power ratio of the two beamformers, a gain factor is defined, which is used to determine the diagonal loading value of the covariance matrix, as well as being a discrimination factor of the post-filtering. The calculation of the threshold of the discrimination factor is also introduced. Experiments in the circular microphone array show that the proposed method performs better in suppressing the interferences without decreasing the speech intelligibility.