Fractional Lower Order Spatial Time-Frequency Blind Source Separation In Infinite Variance Noise Environment

The traditional time-frequency distribution and bind source separation are poor performance in the impulsive noise with α-stable distribution environment. Time-frequency distribution in the presence of impulsive noise is investigated and the new time-frequency distributions which can work in α stable distribution environment are improved. Therefore, the traditional WVD time-frequency distribution are improved based on the stable distribution noise of non-stationary signal, we put forward a kind of fractional lower order pseudo FLO-PWVD time-frequency distribution which can be suitable for α stable distribution noise, and presented the fractional lower order pseudo spatial time-frequency matrix (FLO-MSTFM) concept, and so a new replaced bind source separation algorithm based on fractional lower order spatial time-frequency matrix is proposed, and the algorithm steps are summarized. Computer simulations show that the new algorithm has good performance to separate non-stationary signals in α-stable distribution environment and implement blind extraction of the α stable distribution signal, the MSE is less than TF-UBSS algorithm and MD-BSS algorithm in different α and GSNR.