Underdetermined Blind Source Separation Method Based on a Two-Stage Single-Source Point Screening

Underdetermined blind source separation is a signal processing technique that more suitable for practical applications and aims to separate the signals from mixed signals. The mixing matrix estimation major step in underdetermined separation. Since current methods estimating have disadvantages of insufficient accuracy or weak noise immunity, two-stage single-source point screening combines cosine angle algorithm L1-norm optimization proposed. During first stage, first-stage points are extracted original using algorithm. second based on algorithm, reference then clustered obtain clustering center, which defined as center. In combination with deviation interference eliminated by distance. remaining considered second-stage points, estimation. Experiments simulated speech show proposed method can accurate robust estimation, leading better