Blind Source Separation Algorithm Based on Short-Time Fourier Time-Frequency Analysis

A blind source separation algorithm utilizing short-time Fourier time-frequency analysis, designed for convolutional mixture models to process non-stationary source signals. The implementation involves time-frequency domain transformation, feature extraction using spectral characteristics, and separation via statistical independence criteria.