The Algorithm Research of Blind Source Separation Based on Spatial Time-frequency Distribution for FSK Signal

Abstract

Track circuit is an important guarantees of the railway security, a large number of non-stationary interference and noise induce from the electrified railway block in the process of the FSK signal transmission in the track circuit, seriously affect its testing accuracy and reliability. Aiming at the problem, this paper propose a algorithm of blind source separation (BSS) based on the spatial time-frequency distribution (STFD) and singular value decomposition(SVD). First using SVD on mixed signal to remove noise, then reassign spectrum and pre-white in STFD domain, improve the selection function of single autoterm in STFD matrix, the get unitary matrix by joint diagonal, last estimated out the mixed matrix. The algorithm is not limited with Gaussian source signal, so it has strong applicability. Computer simulation results indicate that the algorithm to effectively separate the non-stationary signals and improve its BSS performance in a low signal-to-noise (SNR) environment.