Direction-of-arrival estimation based on joint diagonalization of matrices in different direct-to-reverberation ratios

Abstract

Direction-of-arrival (DOA) estimation is very important in many applications. The MUltiple SIgnal Classification (MUSIC) algorithm is one of the well-known methods of DOA estimation. However, in a real environment with reverberation, MUSIC's estimation accuracy is degraded. In this paper, we propose an improved MUSIC algorithm by using joint diagonalization of the covariance matrices of two different direct-to-reverberation ratio (DRR) periods corresponding to higher and lower DRR periods in a short time. In addition, we also introduce a method for detecting different DRR periods by using peak-hold processing for our proposed method. Computer simulations verified that the proposed method improves the accuracy of the azimuth in DOA estimation compared with the conventional method for both simulated and actual impulse responses.