Bayesian DOA estimation method using Population Monte Carlo

Abstract

Bayesian maximum a posteriori (BMAP) DOA estimation method has a better performance than MUSIC method at low signal to noise ratio (SNR) and few snapshots. However, it suffers a heavy computational complexity due to multi-dimensional search. Monte Carlo methods such as Markov Chain Monte Carlo (MCMC) method can effectively solve this problem. MCMC method may provide a local optimization for it is difficult to assess when the Markov Chain has reached the stationary state. Population Monte Carlo (PMC) which uses sequential techniques draws a set of particles and provides an unbiased estimate at each iteration. Thus provides a global optimization and can enhance the computational efficiency. In this paper, the PMC method is introduced and used for Bayesian DOA estimation in order to reduce the complexity. Simulation results show that it has better performance than MUSIC method at low SNR or few snapshots. Compared with BMAP, it can reduce the computation and keep high resolution performance at low SNR.