A covariance approximation method for near-field coherent sources localization using uniform linear array

The covariance approximation (CA) multiple signal classification (MUSIC) is a novel near-field direction-of-arrival (DoA) estimation method for uniform linear array. In this paper, we show that the CA-MUSIC suffers from significant performance degeneration caused by coherent sources. The CA-MUSIC with coherent sources generates the image sources (IS), which cannot be distinguished from the real sources. To solve this problem, we propose a CA-based near-field coherent sources localization algorithm, which is robust to the IS effect. The proposed CA algorithm avoids errors caused by coherence between sources using searching radius restriction and zero-forcing MUSIC. Simulation results shows that the proposed CA algorithm offers superior root mean square error (RMSE) performances for near-field coherent sources.