DOA estimation using fast EM algorithm

We study the direction of arrival estimation using expectation-maximization (EM) algorithm. The EM algorithm is a general and popular numerical method for finding maximum likelihood estimates which usually has a simple implementation and stable convergence. However, the computational cost caused by the slow convergence of the EM algorithm is still immense for the direction finding problem. Motivated by componentwise convergence of the EM algorithm, we suggest the use of smaller search spaces after a few iterations. In this way, the overall computational cost can be reduced drastically. An adaptive procedure which determines the search spaces involved in the maximization (M) step is proposed. With numerical experiments we demonstrate the improvement of the computational efficiency by using the proposed algorithm.