DOA Estimation via Acoustic Sensor Array With Non-Orthogonal Elements

A single two-dimensional acoustic vector sensor (AVS) has been widely used to estimate the direction of arrival (DOA) of an acoustic signal. In general, the existing techniques are derived under the assumption that the two velocity sensors are orthogonal. However, the velocity sensors may not be perfectly orthogonal and the angle mismatch will cause the DOA estimation performance to deteriorate. In this paper, we propose an iterative DOA estimation method to solve this angle deviation issue. In particular, we first introduce the deviation parameter into the DOA estimation problem, and quantify the DOA estimation bias based on the two non-orthogonal AVS models. It is shown that the non-orthogonal sensor elements result in an asymptotically biased DOA estimate. Then, a new iterative method is proposed to estimate jointly the DOA of acoustic source and the non-orthogonal deviation matrix (or angle deviation). It is shown that the proposed method can be applied in conjunction with conventional DOA estimation methods to improve their estimation accuracy under non-orthogonal AVS. The effectiveness of the proposed method and its extension is confirmed by Monte Carlo simulation as well as underwater experiment.