Research On Signal Separation Method For Time-frequency Overlapped Signal From Single Channel

Single-channel time-frequency overlapped signal separation is an important research in the communications field in the recent year.For its special receiving mode and the characteristics of the received signal in the time-frequency domain,we cannot separate such a signal separation by the technology of the traditional filtering.The subject is mainly to study the separation of time-frequency overlapped mixed signal composed of two signals which is received by a single sensor.In this context,this paper will focus on the solution of these problems.The main contributions are as follows:1.Research on the selection of wavelet basis in signal separation algorithm based on multi-scale decomposition.First,a model of a signal separation algorithm based on multi-scale decomposition will be constructed,and the characteristics of the required wavelet basic function are obtained by analysis of the algorithm framework.This type of wavelet is regarded as the wavelet basis under the framework of the algorithm.The separation effect of the signal group which is obtained from the method based on the different wavelet will be compared and analyzed.And the separation performance of different signal group which is obtained from the method based on the different wavelet with the different signal-to-noise ratios will be given,and the properest basic wavelet function of the method based on the multi-scale decomposition will be selected.2.Research on the parameter optimization problem of Morlet wavelet in the method based on multi-scale decomposition.By analyzing the characteristics of the Morlet wavelet and the shortages of the wavelet in the signal separation algorithm based on multi-scale decomposition,an adaptive Morlet wavelet will be proposed.To a certain extent,the optimization of the traditional Morlet wavelet center frequency based on the minimum Shannon entropy criterion can solve the problem of the wavelet parameters which is relying on empirical value settings,the operation improve the performance of the wavelet.The optimized Morlet wavelet is used as the wavelet basis of the separation method based on the multi-scale wavelet decomposition,this method will perform on the received time-frequency overlapped mixed signal to obtain several wavelet coefficients,these coefficients will be utilized to construct the virtual multi-dimention observed signal and independent component analysis will be used to complete the signal separation.To a certain extent,The simulaiton shows that the optimized wavelet has been improved.The simulation shows that the operation makes the algorithm more intelligent,and the optimized method is still effective for the separation of multiple sets of mixed signals.3.Research on the amplitude ambiguity correction of the signal separation part of the method based on multi-scale decomposition.First,analyze the natural amplitude uncertainty of the ICA part,and list common methods which are used to eliminate amplitude ambiguity.On the basis of these common methods,a method based on amplitude ambiguity correction will be designed to optimize the signal separation part.To a certain extent,This method can solve the amplitude ambiguity of the signal separation module and improve the algorithm's ability to separate the mixed signal containing amplitude modulated signals.We verify whether the proposed method is effective to separate the mixed signals containing amplitude modulated signals.And the feasibility of the optimization scheme will be proved by comparing the algorithm with the algorithm before improvement,And analyze the influence of mixed vector and frequency overlapped level on the optimized method.