Research On Time-Varying Filtering Algorithm And Its Application In Communication Signal Recognition

With the continuous development of communication technology,the modulation types of communication signals are being more and more complicated.How to quickly and accurately identify the modulation types of communication signals is of great significance in both military and civilian fields.Under low signal-to-noise ratio conditions,noise causes interference to the feature extraction and recognition of communication signals,and seriously affects classification effect of communication signals.It is necessary to preprocess the signal to de-noise.The traditional filtering method can only achieve a single filtering in the time domain or the frequency domain,and could not combine time-varying characteristics of communication signal,and its filtering effect is not ideal.Therefore,how to design suitable time-varying filtering algorithm to filter out noise while retaining signal characteristics has important research significance for improving identification rate of communication signals.For the problem that noise seriously affects the recognition performance of communication signals under low signal-to-noise ratio conditions,this paper designs a time-varying filtering algorithm to filter out noise in the communication signal preprocessing stage,and uses signal recognition classifiers based on convolutional neural networks to achieve the communication signal modulation mode recognition under low signal-to-noise ratio.This article first studies and analyzes the modulation principle and time-frequency feature extraction of communication signals,then introduces relevant basic theories of convolutional neural networks,which supplies theoretical foundation for the subsequent design of communication signal identification algorithms based on convolutional neural networks;The signal time-varying filtering algorithm is studied.A time-varying filtering algorithm based on time-varying frequency extraction and a timevarying filtering algorithm based on time-varying frequency mask are proposed for frequency shift keying intra-class identification and amplitude shift keying,frequency shift keying,phase shift keying inter-class identification of three communication signals.Feature interference;finally,a signal recognition classifier based on convolutional neural network is designed by building,training and analyzing properties of convolutional neural network model.Finally,through the designed communication signal recognition algorithm based on convolutional neural network,the frequency shift keying signal in-class recognition before and after time-varying filtering and the inter-class recognition of amplitude shift keying,frequency shift keying,phase shift keying three kinds of communication signals are simulated and compared and analyzed.Experimental results indicate that the timevarying filtering algorithm devised in paper filters out the interference caused by noise on time-frequency characteristics of the signal with effect,and significantly improves the recognition rate of the modulation type of the communication signal.It has important reference value for the research and development of the communication signal recognition.