宽间距麦克风阵列聚合角度谱空间混叠抑制方法研究  

Spatial aliasing suppression approach for pooled angular spectra of widely spaced microphone arrays

作　　者：陈纯增 许志勇[1] 赵兆[1] 

机构地区：[1]南京理工大学电子工程与光电技术学院,江苏南京210094

出　　处：《南京理工大学学报》2016年第6期700-707,共8页Journal of Nanjing University of Science and Technology

基　　金：国家自然科学基金(61401203;61171167);江苏省自然科学基金(BK20130776)

摘　　要：针对宽间距麦克风阵列在混响环境下的多声源波达方向估计问题,基于声源信号时频稀疏特性研究了1种能有效抑制空间混叠影响的聚合角度谱方法。该方法采用声源能量积累策略和恒虚警率检测技术,仅对频域积累输出中的高值分量进行帧间时域聚合处理,显著降低了聚合角度谱中的空间混叠伪峰高度和数量。仿真结果表明,该方法能够明显增大真实声源谱峰与伪峰之间的高度差,对宽间距阵列声源定位时的空间混叠抑制效果优于现有聚合角度谱方法。A pooled angular spectra method capable of suppressing spatial aliasing effectively is studied based on time-frequency sparsity of acoustic signals to estimate multi-source directions of arrival in reverberant environments for widely spaced microphone arrays. High-valued elements of spectral accumulation outputs are temporally pooled over frames by using the energy accumulation strategy and constant false-alarm rate detection technique. The height and number of spurious peaks of spatial aliasing are significantly lowered in the pooled angular spectra. Simulation results demonstrate that the height difference between true peaks associated with target sources and false ones can be obviously enlarged, and this method outperforms the existing counterparts in terms of spatial aliasing suppression for widely spaced arrays.

关 键 词：麦克风阵列 聚合角度谱 空间混叠 能量积累 恒虚警率检测 时频稀疏 

分 类 号：TN912.3[电子电信—通信与信息系统]