Radio Monitoring Signal Separation Study Based On Cyclic Statistics

For various reasons, such as intensive wireless communication environment, multiple radar target echo, and more generally the presence of noise, etc., the received signal does not only contain the desired signal but also other interfering signals or noises. Worse, interference and noise exit at the same time. How to recover the desired signal best from the mixed signals and how to suppress interference signals to recover the desired signal. This is one of the main direction in signal processing for decades.We specially consider the case of signal spectrum overlap, in this condition, the general separation method become bad. As most of the communication signal can be seen as cyclostationary signals, the cyclic statistics is still likely to make signal separative in the frequency overlap case.This paper studies some known separation algorithms, has an understanding of the signal model and the environment, and simulates them.Through extensive literature review, based on the sign natural gradient algorithm and the second order cyclic statistics, the paper proposes a fast convergence of the cyclic sign natural gradient algorithm. The method uses circular whitening which removes the correlation between the signals in the cyclostationary frequency domain, so that the convergence speed is improved.Further more, a simple blind source separation method based on the properties of forth-order cyclic cumulant and cyclic whitening is proposed. In the condition of two mixed cyclostationary signals, firstly the observed signal matrix is cyclic whitened, then the cyclic autocorrelation matrix become a unit matrix. Thus the separation matrix is turned into a unitary, too. which can be described by one parameter. Then the optimal value of this parameter can be achieved by the judge function based on the characteristic of cyclic statistics, through which the separating matrix can be determined. The analysis of the result figure of the signal separation and crosstalk error show the effectiveness of the proposed method. Besides, the advantage on computation is talked about.