Research And Application Of Radar Emitter Recognition Technology Based On Reinforcement Learning

Radar emitter source individual recognition also known as radar emitter fingerprint identification,refers to the process of receiving electromagnetic signals emitted by radar emitters,extracting fingerprint features,and determining the individual emitter which generates the signal based on existing information.The intelligence level of modern electronic warfare is enhanced with the continuous development of information technology.And at the same time,the continuous update of new system radar emitters makes the signal waveform more complex and changeable.Due to the lack of autonomous learning ability,the traditional identification methods based on template matching have been difficult to meet the needs of the actual battlefield.So,it is a huge challenge for the task of radar emitter source recognition.Therefore,in response to the above problems,this thesis proposes new individual recognition schemes for radar emitter.Starting from the reinforcement learning model,this thesis studies the fingerprint feature selection of individual radar emitters and the design of classifiers.The main works of this thesis are as follows:Firstly,the theoretical basis of radar emitter individual recognition is deeply studied in this thesis.Two kinds of radar transmitter structures are compared and analyzed,and the mechanism of signal generation is studied.Then the mathematical model of radar signals is analyzed and modeled.Finally,two types of intrapulse modulation of radar emitter signals are analyzed,and raised the unintentional modulation is the main cause of fingerprint feature.Secondly,the individual recognition of radar emitter based on reinforcement learning feature selection algorithm is realized.Based on the embedded feature selection algorithm,the feature selection process is achieved by using the reinforcement learning algorithm,and the recognition of the individual radar emitter is realized by extracting fingerprint features.The accuracy of the recognition reached to 99.98%.The feature selection algorithm proposed in this thesis improves the accuracy of individual recognition of radar emitters while reducing the feature dimension of fingerprint features.Finally,the scheme of radar emitter individual recognition based on deep reinforcement learning algorithm is realized in this thesis.DQN,DDQN,and Dueling Network are used as classifiers to apply Markov decision model to the individual recognition task of radar emitters.The feasibility of these models is verified by measured data,and the recognition accuracy is over 99%.