Blind Separation And Modulation Recognition Algorithm For Single Channel Aliased Signals

In the traditional blind source separation task,several signal components of the aliased signal are highly overlapped in the time domain or frequency domain,and the signal is received by the receiver in the form of a single antenna,resulting in the basic failure of some existing signal modulation identification methods such as time-frequency domain.Therefore,it is necessary to consider introducing some optimization links or new methods to deal with the existing problems.In this paper,the aliased signal with Gaussian white noise is received in a single channel under the condition of being completely blind as far as possible,that is,without knowing the parameters of the transmitted signal,and the aliased signal with noise is separated by the designed algorithm to obtain two estimated signals.Finally,the modulation mode of the two signals is identified.This paper mainly carries out the following research:(1)In this paper,an aliasing signal separation method based on multi-resolution singular spectrum combined with improved principal component analysis is proposed.By taking advantage of the time-delay characteristics of the signal itself,the algorithm retains more useful information,expands the single channel signal to multi-channel signal,so as to transform the under determined problem into a well determined problem for processing,and then speeds up the noise reduction process through the improved principal component analysis method.The performance of the aliasing signal separation algorithm is significantly improved compared with similar algorithms,and the performance of the algorithm is verified by simulation experiments.(2)Through the entropy bound estimation method,the separated signal is further denoised,and the circular convolution neural network modulation recognition method with stronger robustness to the modulation signal is constructed to find the characteristics of the estimated signal after blind separation of aliased signals,so as to fully retain the necessary information when inputting a long sequence,so that the neural network can focus more on finding the useful information significantly related to the current output in the input data,Improve output accuracy.Through the horizontal index comparison experiment with similar methods,the performance of the proposed algorithm is compared.Finally,the proposed two algorithms are combined,and the blind separation and recognition scheme is designed in detail according to the actual application scenario.Through the designed aliasing signal measurement method,the algorithm performance in the measured scene is obtained to verify the practical value of the proposed algorithm.