Third-Order Cumulants-Based Method To Estimate 2-D DOAs Of Multiple Sources Using An AVS

Abstract

A spatially co-located acoustic vector sensor has several benefits over the microphone/hydrophone array like small size, and the inherent two-dimensional directivity, which is independent of the frequency, range, and bandwidth of the incoming source signal. In spite of all these advantages, a single acoustic vector sensor can estimate only the maximum two sound sources using conventional second-order statistics based methods. Higher-order statistics are well known for their aperture/degree of freedom extension property in the field of array signal processing. Therefore, this paper proposes a compressive sensing based method by exploiting the third-order cumulants to estimate two-dimensional direction of arrivals of multiple non-Gaussian sources. The proposed method enables an acoustic vector sensor to estimate two-dimensional direction-of-arrivals of maximum five statistical independent, non-Gaussian sources, provided both the elevations and azimuths are distinct for all the given sources. Further, the maximum three two-dimensional direction of arrivals are to be estimated using a single acoustic vector sensor by the proposed method, if all the given sources are lying in a single azimuth (or elevation) plane with distinct elevations (or azimuths).