Detection And Recognition Of Time Frequency Overlapping Interference Signal

In the increasingly complex electromagnetic environment,time-frequency overlapping interference signals are widespread.Communication signals often suffer from the intrusion of interference signals in the same frequency band.It affects the normal reception of communication signals at the receiving end.By detecting and identifying the time-frequency overlapping interference signal in the communication system,the influence of the interference signal can be effectively reduced or avoided.Based on the in-depth study of the detection and identification of time-frequency overlapping interference signals,the problem is decomposed into three successive sub-steps,interference signal detection,interference signal separation,and interference signal classification and identification.Combining the ideas of feature engineering and deep learning,three corresponding models are proposed.The detection and recognition of time-frequency overlapping interference signals are realized effectively.The main work content of this thesis includes the following three aspects.1.An interference signal detection algorithm model based on time-frequency power entropy is proposed,the problem of difficult interference signal detection is firstly studied.When the interference signal power is lower than the communication signal power,the time-frequency power entropy in the three dimensions of time domain,frequency domain,and power domain is used as the input feature of the interference signal detection model.Considering the fact that the intrusive interference signal anomaly signal samples in the information battlefield are often less than the actual situation of the normal signal samples,the Machine learning classification algorithm of One class Support Vector Machine(Oc-SVM)is used as the classifier of the interference signal detection model.The experimental scenario of the interference signal intruding into the communication signal is designed.Simulations are provided to demonstrate to verify the feasibility and effectiveness of the proposed interference signal detection algorithm model.2.A blind separation algorithm model of interference signals based on convolutional auto-encoder is proposed.Aiming at the problem of blind separation of interference signals,the convolutional neural network is used as the basic structure of the auto-encoder,and the single-task auto-encoder model is adopted to separate the interference signal through the difference between the input of the network and the predicted output.When the power of the interference signal is much higher than the power of the communication signal,the single-task auto-encoder separation effect is not good.Therefore the multi-task auto-encoder model is applied to separate the interference signal.An experimental scenario for separating time-frequency overlapping multi-tone interference signals and communication signals is designed.Simulation results show the feasibility and effectiveness of the proposed interference signal blind separation algorithm model.3.An interference signal classification and identification algorithm model based on Gated Recurrent Unit(GRU)is proposed.After the independent interference signal is obtained through the second step above,the interference signal is regarded as a time series for the identification of the interference signal.The neural network constructed by GRU is used to extract the hidden features in the interference time series data,and the recognition result of the interference signal is obtained.Two recognition and classification algorithms based on the traditional feature extraction and the GRU neural network proposed in this thesis are compared and studied.The basic principles of each algorithm are described.The t-distributed Stochastic Neighbor Embedding(t-SNE)algorithm is used to process the data of the penultimate layer of the GRU network,which explains the feasibility of the GRU algorithm.Four kinds of interference signal classification and identification experimental scenarios under different sample conditions are designed,and the performance of the proposed interference signal identification model is simulated and compared.