PESO: Probabilistic evaluation of subspaces orthogonality for wideband DOA estimation

Abstract

This paper introduces a novel direction-of-arrival (DOA) estimation method for the closely related wideband sources. The new method estimates the DOAs accurately by evaluating the probability relation between the signal and the noise subspaces of multiple frequency components of the sources using supervised singular value decomposition (SupSVD) and likelihood mean shift (LMS) methods. Also, the introduced method uses the selective criteria of the reference frequency in minimum noisy subband-TOPS (MNS-TOPS) which is an improved version of test of orthogonality of projected subspaces (TOPS) method. This reference frequency is used as the primary set of interest and other sub-bands are used as supervised set to accurately extract the DOAs from a very noisy reception of the signals. The performance of the introduced method compared with well-known methods such as incoherent signal subspace method (ISSM), TOPS, weighted squared-TOPS (WS-TOPS), and MNS-TOPS. The simulations show that the new method outclasses all other methods in all ranges of SNR values. The newly introduced method tested with the lowest ever used number of snapshots (12) and very low SNR values ( $$<-$$ 10 dB). The new method estimates the exact DOAs at the lowest computational cost while the conventional methods can not produce accurate results.