Design And Implementation Of Radar Radiation Source Intelligent Identification Platform Based On Deep Learning

Radar radiation source identification is an important part of radar information countermeasures,and plays a decisive role in winning or losing electronic warfare.With the increasingly complex electromagnetic environment of the battlefield and the development of new system radar technology,the form of radar radiation source signals has become increasingly complex and unknown,and presents a "big data" trend,which poses serious challenges to traditional radar radiation source identification methods.The development of artificial intelligence has brought new opportunities to the development of radar radiation source recognition technology,and the deepening of the deep learning theory has also provided a new research direction for recognition algorithms.In this paper,a radar radiation source signal recognition platform is designed and implemented in conjunction with deep learning theory research.A radar radiation source recognition algorithm based on deep learning network is proposed and tested and verified on the platform.The platform can realize the processing of simulated data and measured data of radiation source signals,and can also complete the comparative test and verification of multiple radiation source identification algorithms.The research in this article is as follows:1.The architecture of the radar radiation source recognition platform based on deep learning is proposed.The composition and function of each module are introduced from the software and hardware levels of the system.The detailed process of radar radiation source signal processing in the platform is given in detail.2.The research analyzes the form of the radar radiation source signal,designs the radar radiation source signal acquisition and transmission module,and introduces the data processing process of radiation source signal generation,collection,transmission and storage.3.Several common deep learning architectures are studied,and a deep learning network recognition algorithm based on multi-layer perceptron is proposed.Using short-time Fourier transform to transform the signal into the time-frequency domain for analysis and processing,extracting the time-frequency domain characteristics of the signal can improve the recognition rate of the radiation source recognition algorithm.For the problem that the deep network training takes a long time,this paper uses principal component analysis to reduce the dimensionality of the signal.At the same time,the Dropout function is added to the network structure to randomly propose the connection of some neurons,which can effectively alleviate the occurrence of deep network overfitting.To improve the training efficiency of deep learning networks.After multiple iterations the network reaches the set threshold and the network training is stopped and the multi-layer perceptron network model is saved,and the test data set is used to test the radiation source recognition performance.4.The test verification of the deep learning radar radiation source recognition platform based on the multi-layer perceptron is completed.First collect the radar source sample data,and then transfer the collected source sample data to the system platform,read the collected source signal and do time-frequency transformation and dimensionality reduction processing,and call the third-layer trained multi-layer perceptron network The model,read the test data,start the upper computer of the radar radiation source recognition platform to obtain the signal recognition result.