An effective doa estimation by exploring the spatial sparse representation of the inter-sensor data ratio model

Abstract

This paper investigates speaker direction of arrival (DOA) estimation using a single acoustic vector sensor (AVS). With the definition of the inter-sensor data ratio (ISDR) in the time-frequency (TF) domain and the use of the high local signal-to-noise ratio (HLSNR) TF points, an effective ISDR data model is derived, which determines the relationship between the ISDR and the AVS manifold vector. With the spatial sparse representation of the ISDR data, the DOA estimation is formulated by recovering the sparse matrix and locating the peak of the power spectrum of the reconstructed sparse matrix. Preliminary experimental results using simulations and real AVS recordings show that the proposed DOA estimation method is able to achieve high elevation and azimuth estimation accuracy for all angles when the SNR is above 10dB, avoiding the spatial aliasing problem and suppressing the adverse impact of the room reverberation. It is expected that the proposed DOA estimation method may find wide applications in portable devices due to its small compact physical size and superior performance.