BLIND SOURCE SEPARATION METHOD AND SYSTEM BASED ON SEPARATION MATRIX INITIALIZATION FREQUENCY POINT SELECTION

摘要

Disclosed are a blind source separation method and system based on separation matrix initialization frequency point selection. The method comprises: obtaining an audio signal to be separated and performing Fourier transform; performing 1-time ICA iteration on frequency points in a frequency domain range in which spatial aliasing will not occur to obtain a separation matrix, and estimating DOA information of each source signal; performing frequency point selection on each frequency point in the frequency domain range according to a mixed signal covariance matrix, and classifying the frequency point into a preliminarily-selected frequency point set; performing ICA iteration on preliminarily-selected frequency points, and performing initialization by using the DOA information of the source signals to obtain an initial separation matrix; then performing ICA iteration on the preliminarily-selected frequency points by using the initial separation matrix to obtain a separation matrix of the preliminarily-selected frequency points, and re-estimating DOA information of the source signals; constructing a separation matrix of unselected frequency points on the basis of the re-estimated DOA information; and performing inverse Fourier transform according to the separation matrixes of all the frequency points, and performing reconstruction to obtain separated signals. According to the present disclosure, by initializing a separation matrix, the convergence rate of an algorithm is increased, and the separation performance is improved.