Recognition Algorithm For The Four Kinds Of Interference Signals

Under the condition of informationized warfare,ultrashort wave communication systems are facing the threat of increasingly serious communication interference.Accurate and efficient identification of various communication interferences is an urgent problem to be solved.Faced with this complex and challenging research topic,the traditional pattern recognition method has the disadvantages of blindness,randomness and vulnerability to noise caused by the need to manually extract a large number of features.Therefore,this thesis proposes a interference signal recognition algorithm based on time-frequency analysis and convolutional neural network.In this thesis,four kinds of typical communication interference signals are used: single tone interference,multi-tone interference,linear frequency sweep interference,and noise frequency modulation interference.From the instantaneous characteristics of communications interference signals,which are time-frequency domain distribution characteristics,combining with the advantages of automatically extracting image characteristics of powerful convolutional neural network(CNN),this thesis proposes a new type of interference recognition algorithm.This thesis firstly analyzes the time domain and frequency domain characteristics of binary phase shift keying(BPSK)communication signals and the above four types of interference signals.Next based on the smoothed pseudo Wigner-Ville distribution(SPWVD)time-frequency analysis method,an improved method is proposed and the improvement effect is verified.The time-domain analysis of the mixed signal of the communication signal and the interference signal is carried out by using the threshold-based SPWVD algorithm proposed in this thesis.The characteristics of the time-frequency domain distribution image are expounded.Then according to the principle of convolutional neural network,a interference recognition algorithm based on time-frequency analysis and convolutional neural network is proposed.The method uses the time-frequency domain distribution image of the above-mentioned mixed signal as the training samples and test samples.The network framework consists of basic components of convolutional neural networks,this algorithm can automatically extract deep features of the image without manually defining the extracted features.In this thesis,the parameters and methods of algorithm training samples and test samples are given.The process of algorithm model training is expounded,and the simulation results and performance analysis are listed.The simulation results show that when the signal-to-interference-and-noise-ratio(SINR)reaches 0dB,the accuracy of the four types of interference recognition exceeds 99%,which indicates that the proposed interference recognition algorithm can preeminently identifies the four typical communication interference signals.