Emitter Signal Recognition Based On VMD And Deep Learning

With the development of electronic countermeasures technology,the electromagnetic environment has become more complicated,which also puts forward higher requirements for electronic reconnaissance and interference.Emitter recognition is a key content in the field of electronic warfare technology.In order to meet higher requirements for recognition rate,noise resistance and timeliness,based on deep learning,the modulation recognition method of emitter signals is studied.First,in order to solve the problem of poor anti-noise performance of artificially extracted features and poor timeliness of time-frequency images combined with deep learning methods,an algorithm based on 1-Dimensional Multi-channel Convolutional Neural Network(1DMCNN)is proposed.Modal decomposition is mainly used to obtain multi-channel one-dimensional signals.Variational Mode Decomposition(VMD)has a good decomposition effect on the emitter signals.The fast Fourier transform makes the selected features not affected by changes in frequency.The simulation results show that the recognition rate of the 1DM-Le Net-5 model is 70% when the SNR is-6d B.The recognition rate of the 1DM-Alex Net-8 model is 94%.Secondly,as the number of layers of the CNN increase,the timeliness will be greatly reduced.In order to solve this problem,the Le Net-5 model was optimized using Batch Normalization(BN)and Dropout.The experimental results show that when the SNR is higher than-6d B,the recognition rate of the six emitter signals remains above 93%;when the SNR drops to-8d B,the recognition rate still reaches 90%.Compared with the benchmark model,the recognition rate is increased by nearly 20%,and the network scale is smaller;compared with manual feature extraction combined with machine learning methods,it has a higher recognition rate;compared with time-frequency images combined with deep learning methods,timeliness is greatly improved.In summary,this paper proposes an emitter signal recognition method based on VMD and 1DMCNN.Under the condition of meeting a certain recognition rate,it can effectively solve the problem of anti-noise and timeliness in emitter recognition.At the same time,in view of the development of deep learning in the fields of speech,image,etc.,applying it to the recognition of emitter signals has certain research value.Figure 37;Table 22;Reference 62...