关键词: 盲源分离, 联合对角化, 奇异值分解

Abstract: Aimad at the blind separation problem with time structure, this paper proposes a joint diagonalization algorithm based on two positive definite matrices. Firstly, two positive definite matrices are constructed by using several different time delay statistics. Next, a new algorithm is proposed to realize the joint diagonalization of the two positive definite matrices. Then separation matrix can be obtained. Finally, the source signals can be estimated. The proposed algorithm overcomes the shortcomings of the existing algorithms： the heavy computation load caused by the joint diagonalization of multiple matrices and the low separation accuracy caused by the single matrix. Simulation results show that the proposed algorithm is superior to other comparison algorithms in the presence or absence of noise.

Key words: blind source separation, joint diagonalization, singular value decomposition