Number Estimation And Separation Of Single Channel Mixed Signal With Same Frequency

Today,with the rapid advancement of the global informatization process,singlechannel reception of co-frequency mixed signals is becoming more and more common.Due to the ill-conditioned problem of only one received signal for a single channel,the traditional multi-channel mixed signal separation technology cannot be directly applied,so it is necessary to carry out corresponding theoretical research and algorithm implementation.Through the in-depth understanding of the existing single-channel mixed signal processing technology at home and abroad,two major difficulties in the current singlechannel processing technology are summarized:(1)the number estimation of source signals;(2)the recovery of each source signal.In order to find a single-channel mixedsignal processing technology with wider applicability and better effect,this thesis conducts research around these two difficulties.The main contributions are as follows:1.The simulation verifies some existing single-channel source number estimation algorithms and blind separation algorithms.Among them,the number estimation algorithm includes: single-channel source number estimation algorithm based on rearrangement HanKel matrix,number estimation algorithm based on eigenmode decomposition combined with diagonal loading technology,etc.It is found that these algorithms are only applicable under certain conditions,it is completely invalid in other application scenarios;single-channel mixed-signal blind separation algorithms include:blind separation algorithm based on virtual multi-channel,firsthand separation algorithm based on VMD decomposition,etc.These two types of algorithms work better under simple signal sources,but it does not perform well in complex digital signal environments,and its application scenarios are narrow.2.A number estimation algorithm based on db2 wavelet transform is proposed.The method is mainly through the combination of discrete wavelet transform and continuous wavelet transform.Firstly,multiple continuous db2 wavelet transforms are performed on the mixed signal and the amplitude is extracted;secondly,the power spectrum estimation is obtained by using the Welch algorithm;thirdly,perform wavelet packet decomposition and reconstruction on the estimated power spectral lines to reduce burrs and achieve the goal of removing useless information;finally,the repeated spectral lines are removed by means of a sliding window,and the remaining spectral lines are counted to obtain an estimate of the number of source signals.Simulation experiments show that the algorithm is effective and widely applicable.3.Using Morlet multi-scale wavelet decomposition and nc-FastICA algorithm to achieve blind separation of single-channel mixed signals.This method performs Morlet wavelet transform on the single-channel mixed signal on the optimal scale sequence,in which the optimal scale sequence is obtained in multiple sets of scale sequences through the minimum collision entropy algorithm,and then obtains multiple sets of wavelet transform coefficient groups,that is,multi-dimensional signals.Then separate the independent components through the nc-FastICA algorithm,and finally get the estimation of each source signal after screening.Simulation results show that the algorithm has good anti-noise performance and has certain engineering applicability.