Research Of DOA Estimation Of Acoustic Sources In Long Line Array Based On Subspace Intersection Method

The source bearing , or direction of arrival(DOA), estimation is an important aspect in array signal processing. Using large aperture array, the sonar performance can be improved, and the detecting distance from the target can be increased, but the effect of the multipath nature of sound propagation, or (multimode wave), in ocean on DOA estimation has to be considered as the increase of array aperture. The DOA estimation based on subspace intersection method(SI) can effectively suppresses the effect of multipath nature on DOA estimation in large aperture array. Comparing with the matched field processing method, SI needs smaller amount of calculation, but its robustness is very poor. In this paper, an algorithm of SI with better robustness is expected through the improvement of SI.We first analyze the effect of multipath nature on source bearing estimation of the conventional beamforming method in large aperture array on DOA estimation. Then, the subspace intersection method and limited square subspace intersection method (LSSI) are reviewed. The theoretical analysis and the results of numerical simulation show that the spatial spectrum peaks of DOA estimation method based on plane wave mode usually split or shift because of the effect of multipath nature of sound propagation in the ocean, and the more the bearing is close to the endfire, the more this phenomena is obvious. SI suppresses the effect of multimode nature on DOA estimation effectively, but the robustness of SI is not good and the performance of DOA estimation depends on source number estimation(SNE) significantly. LSSI improved the robustness of SI, but still depends on SNE.In this paper, we deeply analyze the robustness of SI and come to the conclusion that the essential reason of the bad robustness of SI is the modal array response matrix (ARM) isn't full-ranked. Then the subspace intersection method of DOA estimation based on white noise loading approach (WASI) is presented. This method loads white noise on the ARM in order to get the full-ranked ARM, so it can improve the robustness of SI obviously. The results of numerical simulation show that WASI has better robustness and performance of DOA estimation than SI. Furthermore, the effect of white noise loading approach on DOA estimation of WASI is analyzed and found to be very small by using a specially designed imaginary experiment. In the sense of DOA estimation, the performance of WASI is very closed to that of SI.In order to improve the robustness of SI and solve the problem of SI depending on SNE seriously, a new subspace intersection method for DOA estimation based on global optimization (GOSI) is presented. GOSI sets up a model based on the joint searching of the source number and bearings. Through the two-dimensional searching of source number and bearings, GOSI improves the robustness of SI effectively and releases the depending on SNE. The results of numerical simulation show that 1) Comparing with the old source number estimation method, GOSI has better performance of SNE. 2) Comparing with the primary SI method, GOSI has better performance of DOA estimation and robustness.