面向多声源的压缩感知麦克风阵列的波达方向估计

(厦门大学 海洋与地球学院,水声通信与海洋信息技术教育部重点实验室,福建 厦门 361102)

压缩感知; 麦克风阵列; 多声源; 波达方向

Direction of Arrival Estimation of Compressed Sensing Microphone Arrays for Multiple Sound Sources
ZHANG Yudong,HUANG Huixiang,TONG Feng*

(School of Sciences,Jimei University,Xiamen 361021,China)

compressed sensing(CS); microphone array; multiple sound sources; direction of arrival(DOA)

DOI: 10.6043/j.issn.0438-0479.201708013

备注

在语音识别、说话人识别等语音交互应用领域中,麦克风阵列常常工作于多声源工作场景,因而需要更高的波达方向(DOA)估计分辨性能.压缩感知(CS)的DOA估计算法可将声源定位的问题转化成稀疏信号的重构问题,进而提高在高混响、低信噪比环境下的DOA估计性能.基于这一思想,将CS方法应用于多声源方位估计.考虑到传统的基于CS的DOA估计算法利用实测声源传输响应作为混合矩阵时,会因噪声的存在而导致多声源条件下的匹配程度下降,提出了利用基于阵列各阵元之间时延关系所生成的不同方位的声源传输响应来构造CS混合矩阵,即构造房间冲激响应CS(CRR-CS)的DOA估计算法,从而实现多声源的DOA稀疏恢复.通过实验验证了该方法优于传统方法,能更好地实现定位.

In applications of voice recognition,speaker recognition and other voice interaction,microphone arrays often work in multiple sound sources scenes.Therefore microphone arrays need higher resolution performance of direction of arrival(DOA)estimation.Compressed sensing-direction of arrival(CS-DOA)transforms the problems of sound source localization into the reconstruction of sparse signals,thus can improve the performance of DOA estimation in high reverberation and low SNR environment.In this paper,the compressed sensing is applied to location estimation of multiple sound sources,the traditional CS-DOA method using the mea-sured sound sources transmission response as a hybrid matrix,which will result in a decrease in the degree of matching under multiple sound sources due to the noise This article constructs the compression-sensing hybrid matrix based on the different azimuth sound sources transmission response generated by the array-element delay relation,which constructs the room impulse response(constructed room response,CRR),so as to achieves multiple sound sources DOA sparse recovery.Finally experimental results are given to verify the effectiveness of the proposed method.