Zeinab Shamaee - dept. of Electrical Engineering University of Isfahan Isfahan, Iran
Mohsen Mivehchy - dept. of Electrical Engineering University of Isfahan Isfahan, Iran

Radar signal recognition is an important issue forintelligence receiver at the concept of cognitive electronic warfare.The presence of jamming is neglected at most of the conventionalrecognition and classification methods. Also many of them useimage features but intrinsic signal characteristic is not considered.Since, in this paper, a novel framework of bidirectional longshort term memory (Bi-LSTM) algorithm is proposed for radarsignal recognition and classification in an environment along withthe influence of additive white Gaussian noise and barrage noisejamming.The proposed method not only use intrinsic features but alsotake advantage of deep classifier to enhance the radar signalrecognition. This method has three stages to achieve appropriateresults: a transform pre-processing, feature extraction part andBi-LSTM network. The simulation results indicate the efficiencyof the proposed method, which can identify radar signals in lowSJRs. The evaluation results indicate that the proposed methodoutperforms the SVM baseline in terms of precision.

Bi-LSTM, radar signal processing, electronic warfare, intelligence, signal classification