基于聚焦信号子空间估计导向矢量的干扰声源抑制方法  

Suppression Method of the Interference Sound Sources by Estimated Steering Vector Based on the Focusing Signal Subspace

作　　者：周静 鲍长春 张旭 ZHOU Jing;BAO Chang-chun;ZHANG Xu(Speech and Audio Signal Processing Laboratory,Faculty of Information Technology,Beijing University of Technology,Beijing 100124,China)

机构地区：[1]北京工业大学信息学部语音与音频信号处理实验室,北京100124

出　　处：《电子学报》2023年第1期76-85,共10页Acta Electronica Sinica

基　　金：国家自然科学基金(No.61831019)。

摘　　要：针对最小方差无失真响应(Minimum Variance Distortionless Response,MVDR)波束形成器对导向矢量失配较敏感的问题,本文提出了一种有效的干扰声源抑制方法 .该方法首先将语音信号的频带划分为多个子带,通过聚焦信号子空间方法估计各子带的声源到达方向(Direction of Arrival,DOA),并采用统计直方图估计各声源的初始DOA;其次,为了减小导向矢量失配,利用声源的空间稀疏性,通过Capon功率构建目标声源导向矢量估计的代价函数,约束目标声源导向矢量远离干扰声源空间;最后,根据估计的导向矢量,估计干扰声源加噪声协方差矩阵,以获得MVDR波束形成器的权重.基于TIMIT语料库的实验结果证明,提出的干扰声源抑制方法的输出信干噪比(SINR)及语音质量感知评价(PESQ)优于参考方法,具有更佳的抗导向矢量失配性能.Based on the problem that the minimum variance distortionless response(MVDR) beamformer is very sensitive to the mismatch of the steering vector, an effective method of suppressing the interference sound sources is proposed in this paper. First, the bandwidth of speech signal is divided into multiple sub-bands, and the direction of arrival(DOA) of sound sources at each sub-band is estimated by the focusing signal subspace method. Specially, the initial DOA of each sound source is estimated via statistical histogram. Second, in order to reduce the mismatch of the steering vector, based on the spatial sparsity of sound sources, the cost function used for the steering vector estimation of the target sound source is constructed by Capon power so that the steering vector of the target sound source is constrained away from the space of interference sound sources. Finally, the covariance matrix of interference sound source plus noise is estimated based on the estimated steering vector for obtaining the weights of the MVDR beamformer. The experimental results on the TIMIT corpus show that the proposed method outperforms the reference methods on the tests of the output signal to interference-plusnoise ratio(SINR) and the perceptual evaluation of speech quality(PESQ) and has a better performance for preventing the mismatch of the steering vector.

关 键 词：语音增强 麦克风阵列 波束形成 聚焦信号子空间 最小方差无失真响应 

分 类 号：TN912[电子电信—通信与信息系统]