Sparse Direction of Arrival Estimation using Real Acoustic Data

Sparse direction of arrival (DOA) estimation has become highly popular especially after the development of compressive sensing (CS) theory. For hardware and/or software efficiency, it is highly common to undersample sensor array measurements using measurement matrices, and perform the DOA estimation on the compressed sensor array data. This paper demonstrates a comparative analysis among sparse DOA estimation, Multiple Signal Classification (MUSIC) and Capon's beamformer using a real acoustic data corpus. Although it operates on the compressed data, the sparse DOA estimation algorithm outperforms MUSIC and Capon's beamformer when an adaptive measurement matrix is used. The findings of this study verify the feasibility of sparse DOA estimation algorithms in practical applications.