A signel-channel blind signal separation method based on signal type difference for time-frequency overlapped signal

Performing blind signal separation (BSS) from the domain in which source signals have difference characteristics is an efficient method to recover sources from the recorded mixture signal(s) in a complex electromagnetic environment, especially for single-channel BSS (SCBSS) problem. This paper proposes a novel time-frequency overlapped signal separation method to rapidly divide digital communication signal and pulse repetition interval (PRI) radar signal based on signal type difference characteristics for the single-channel separation scenarios. Firstly, a robust PRI estimation is applied to single-channel record to identity the sample number in one PRI based on the power difference of two types of original signals and the periodic characteristics of radar signal. Then, we transfer the one-dimensional observation mixed signal record into a multidimensional observation matrix, by setting its column number equals to the estimated sample number in one PRI. After that, we perform singular value decomposition (SVD) on the multi-dimensional observation matrix, and then map the obtained the eigenvalues and eigenvectors into original observation space, respectively, as so achieve the aim of original separation. A series of simulation experiments are carried out, which indicate our method can achieve satisfactory separation results with correlation coefficient noless than 0.95 and Signal-to-Interference Ratio (SIR) more than 28dB when signal to noise ratio (SNR) no less than 10dB. Additionally, it demonstrates performance of our method is better than that of typical separation-FastICA/JADE-and it shows that our method not sensitive to the frequency overlap level (FOL) of the source signal, even FOL as high as 50%, it still can get high-precision separation results with SIR nearly achieve 40dB when SNR ≥ 20dB.