The Research Of Blind Source Separation Algorithm For Single-channel Communication Signals

Due to the rapid growth of global demand for communication services, and the sharp increase in the number of wireless communication station, the development of modern communication technology and national defense science need to face the challenges below. The issue contains more complicated electromagnetic communication environment, less available spectrum resources and more types of interference noise.Single-channel time-frequency signal aliasing widely exists in modern communications environment such as military electronic surveillance, radio frequency spectrum monitoring and emergency communications. For solving the separation of single-channel time-frequency aliasing signals, the traditional blind source separation algorithm using array signal processing failed as well as the methods of filtering in time, frequency, spatial and the strategies in coding domain also are not applicable. So how to effectively implement blind source separation of single-channel time-frequency aliasing signals in complex wireless communication environment has important significance.This thesis basing on the analysis and summary of the basic principles of blind source separation, research on the cyclic spectral domain detachable theory of communication signals and construct the cyclic spectral domain filter to achieve single channel blind source separation.Researches found that the cyclostationarity can reveal signals’essential characteristics. According to estimation methods and characteristics of cyclic spectrum, an estimation algorithm of cyclic spectrum based on the time-varying combination ARV model is derived. In the algorithm, the time-varying combination ARV model is constructed for communication signals, the linear non-stationary problem is transformed to linear time-invariant problem through base time function. Then, the covariance matrix and spectral correlation theory are used to estimate the signals’cyclic spectrum. Meanwhile, analyzing the cyclic spectrum of common modulation signals to test and verify the effectiveness of this algorithm. The algorithm is applied to estimate the carrier frequency and symbol period of modulation signals, and the frequency-shifting amount of cyclic spectral domain filter can be determined.Theoretical analysis shows that the communication signals in cyclic spectral domain has independence and sparsity, blind source separation for single-channel communication signals can be achieved by filtering method. For this reason, on the basis of the structure and principles of wiener filter and blind adaptive frequency-shifting filter, one apply the determined frequency-shifting amount to the linear frequency shift conjugate linear filter to get a single-channel blind separation. The method can separate out two source signals at one time. At last, time-frequency overlapped QPSK/BPSK signal adding white noise is simulated in the MATLAB environment. The result indicates that the two source signals can be separated effectively with proposed method.