Discrete spectrum feature analysis method and device for multiple space-time source signals

The invention discloses a discrete spectrum feature analysis method and device for a multi-space-time source signal. The method comprises the following steps: acquiring a sound signal sample matrix X of M sound signal segments from different sources; and performing noise reduction processing from the X by using a wavelet threshold function algorithm, adjusting a filtering parameter of the wavelet threshold function algorithm according to a correlation coefficient between each sound signal segment in the X and a noise reduction result thereof, and performing noise reduction processing again until the number of iterations is equal to a preset maximum number of iterations or an absolute value of the correlation coefficient is smaller than a preset value, finally obtaining a noise reduction multi-space-time source signal matrix D formed by M noise reduction processing results; performing centralization and whitening processing on the D to obtain a whitening matrix; and performing discrete Fourier transform on the