Improving the performance of GA–ML DOA estimator with a resampling scheme

Abstract

The maximum likelihood (ML) direction of arrival (DOA) estimator computed by genetic algorithm (GA) for the exact global solution gives a superior performance compared to other methods. In this paper, we present a resampling-based scheme to improve its ability to resolve closely spaced sources, and to enhance its global convergence. For this purpose, multiple GA–ML estimators are constructed in a parallel manner based on resampling of a single data set, then those estimates are involved into a competition, and successful results are selected and combined to generate a more accurate estimate. Numerical studies demonstrate that the proposed scheme provides less DOA estimation root-mean-squared error (RMSE), higher source resolution probability and lower resolution threshold signal-to-noise ratio (SNR) than some popular approaches including GA–ML; and this technique is not sensitive to the array geometry, source correlation, and etc.