Research On Multi-channel Blind Signal Separation Technology Of AIS

In this dissertation,for solving the problem of satellite receiving AIS signal aliasing,a blind signal separation processing method of multi-channel aliasing signal based on the research of Automatic Identification System(AIS)is discussed.The main contents of this thesis include:(1)A novel blind source extraction algorithm based on cumulative autocorrelation is proposed to solve the problem of AIS real signal separation.In the proposed algorithm,the cumulative autocorrelation of the signal is used as the objective function and an modified artificial bee colony algorithm is used for its optimization.A separated signal component is removed using the deflation method,and the entire source signal can be successfully removed by repeating the separation process.Simulation results show that the algorithm can effectively realize the blind separation of mixed satellite-based AIS signals with greater separation precision and lower computation costs than other algorithms.(2)A noise-suppressed complex-valued FastICA algorithm is proposed by using the method of spherical coordinate transformation to generate a dimension-increasing matrix to increase the dimension of the mixed signal.Then,whiten the new mixed signal with discarding the noise.The blind separation of the AIS signals are achieved by using the noisy complex FastICA algorithm.Simulation results show that increasing the dimension of mixed signal by spherical coordinate transform can effectively reduce the influence of noise.And has a better noise suppression performance than that of the way random increase the mixed signal dimension.Compared with traditional complex FastICA algorithm and complex kurtosis maximization algorithm,this algorithm can effectively reduce the influence of noise.(3)An equivariant adaptive source separation via independence(EASI)algorithm based on adaptive step learning index is proposed for blind source separation.The step learning index is derived from the non-Gaussian property of the output signal and the step size of EASI algorithm is exponentially decayed with the increase of the step learning index.In the initial stage,the step size is large.Then the steady-state error decreases with the decrease of step size.The batch processing and adaptive processing of the above algorithm are simulated respectively.Experimental simulation results show that the convergence rate of the proposed algorithm is faster than that of traditional EASI algorithm,and the steady-state error of the adaptive processing of the proposed algorithm is less than that of adaptive processing of traditional EASI algorithm.