采用压缩感知的麦克风阵列远场声源方位估计  

Far Field Compressed Sensing Microphone Array DOA Estimation

作　　者：曾帆 黄惠祥 童峰[1] ZENG Fan;HUANG Huixiang;TONG Feng(Key Laboratory of Underwater Acoustic Comnmnication and Marine Information Technology of the Ministry of Education,Xiamen Univeristy,Xiamen 361100,China)

机构地区：[1]厦门大学水声通信与海洋信息技术教育部重点实验室,福建厦门361100

出　　处：《兵器装备工程学报》2018年第5期134-138,共5页Journal of Ordnance Equipment Engineering

基　　金：国家自然科学基金(11574258);福建省高校产学合作项目(2015H6019);福建省中青年教师教育科研项目(JAS170012)

摘　　要：传统CS-DOA算法采用事先测量获得的声传播冲激响应组成稀疏恢复方程中的混合矩阵,远场条件下测量冲激响应误差增大,导致性能下降。针对该问题,将房间冲激响应(RIP)分解为时延部分和混响部分,并在稀疏恢复方程中将混响部分移至方向矢量项,表明:通过频点叠加的方式,在远场条件下可直接利用阵列空间关系构造冲激响应组成混合矩阵,进行远场压缩感知声源方位估计。实验结果表明:与相位变换加权的可控响应功率和延时-累加定位算法相比,在远场条件下该算法具有更高的方位估计分辨率。The traditional CS-DOA algorithm needs to measure the acoustic propagation impulse response in advance to act as mixture matrix of sparse recovery equation, which leads to substantial performance degradation under far field as the error of response measurement increases. In this paper, the room impulse response is firstly decomposed into the delay part and the reverberation part, then the latter is moved to the orientation vector in the sparse recovery equation. By the way of frequency domain accumulation, we can use the array spatial relationship to directly construct the impulse response to make up mixture matrix, thus enable far field compressed sensing DOA. The experimental results showed that, compared with steered lesponse power-phase transform and Delayed-sum algorithm, the proposed algorithm has a higher DOA resolution performance under the far field condition.

关 键 词：压缩感知 麦克风阵列 方位估计 远场 

分 类 号：TN912.3[电子电信—通信与信息系统]