Research On The Application Of Deep Learning In Communication Signal Modulation Recognition

Signal modulation recognition has important applications,which is widely used in electronic warfare,communication reconnaissance,radar reconnaissance,radio spectrum monitoring and management,and link adaptation.Nowadays the channel environments are becoming more complex and the signal modulation methods are increasing.How to efficiently and accurately complete the signal modulation recognition in the complex environment is a difficult problem to be solved in various fields of communication.Based on the non-cooperative communication of an electronic countermeasure force,this article aims at the problems of low recognition performance and high computational complexity of traditional algorithms in signal modulation recognition,research the key technology of signal modulation recognition in AWGN channel and Rayleigh fading channel based on deep learning.The work and results are as follows:(1)The signal constellation image is transformed into the signal characteristic gray image by using the characteristic gray-scale image generation algorithm,Applying the key technology of image classification to the field of signal modulation recognition provides new ideas for the research of signal modulation recognition methods..(2)Research the communication signal modulation recognition method based on the Inception-V4 model under the AWGN channel.Through simulation experiments and comparative analysis,under the AWGN channel,using the feature gray map as the input of the deep learning model,Compared with traditional methods,the modulation signal recognition process has better anti-noise ability and higher algorithm robustness.When the SNR is greater than 2.5dB,the recognition accuracy of the seven modulation signals is maintained above 90%.In terms of signal modulation type recognition accuracy,the performance of the Inception-V4 network model is better than the Alex Net network model and the Inception Resnet V2-TA model.(3)Research the communication signal modulation recognition method based on the Inception-V4 model when the fading coefficient(gain)in the Rayleigh fading channel is known and unknown.Through simulation experiments and comparative analysis,the Rayleigh fading coefficient is known in the channel Based on the Inception-V4 model,the modulation type recognition of the modulated signal under the low SNR environment still has high accuracy and robustness.In channels with unknown fading coefficients,OQPSK modulation and QPSK modulation are easy to make mistakes,and the overall recognition accuracy is not as good as channels with known fading coefficients.The research results of this paper can provide application reference for the communication signal modulation recognition system based on deep learning model to improve the modulation recognition accuracy and reduce the model complexity in complex channel environment.