Blind Separation of Closely-Spaced Multipath Signals Using an STFT-MUSIC Algorithm

Most of existing blind source separation (BSS) algorithms are greatly restricted by multipath effect and complex noise in practical wireless communication systems, especially when multipath sources are spatially close. To solve the above challenging issue, this paper proposes a short-time Fourier transform (STFT) based BSS algorithm for separating closely-spaced multipath signals in the presence of impulsive and Gaussian noise. Specifically, we develop an effective strategy to select appropriate auto-source time-frequency (TF) points via principal component analysis, and then the detected auto-source TF points are clustered into several TF partitions for the formulation of spatial time-frequency distribution (STFD) matrices, based on which direction-of-arrival (DOA) estimation of the sources in each partition is separately performed using an STFT-based multiple signal classification (STFT-MUSIC) algorithm. Next, the mixing matrix is precisely estimated based on the DOAs of sources due to available array manifold. Finally, multipath sources are separated by implementing the STFD-BSS algorithm. Numerical simulations verify the superiority of the proposed algorithm in the separation of closely-spaced multipath signals compared to some state-of-the-art ones.