Research On LPI Radar Waveform Recongnition Algorithm Based On Joint Time-Frequency Analysis And Ambiguity Function

Low Probability of Intercept(LPI)radar is widely used because of its wide bandwidth and low power.But LPI radar signals can avoid the detection of non-cooperative receivers.The correct identification of intercepted radar signals has a great impact on the battle situation.Therefore,it is important to study the identification of LPI radar signals.However,with the use of more kinds of LPI radar signals,the previous methods of identifying several conventional signals can not meet the requirements,so the research of more kinds of signal recognition methods has the important practical value.In this paper,eight and twelve kinds of radar signal recognition are discussed.The recognition methods based on joint time-frequency analysis(JTFA)and ambiguity function(AF)are proposed respectively.It has high recognition rate under the low signal-to-noise ratio.The main contents are as follows:Firstly,the feature extraction method based on JTFA is studied.It is a key part of the recognition system and has a significant impact on the results.In order to distinguish radar signals,four types of features are extracted,i.e.,time features,frequency features,features based on Wigner-Ville Distribution(WVD)and Choi-Williams Distribution(CWD).In order to distinguish LFM from Frank and P3 codes,a new WVD feature is proposed,and two new CWD features are proposed to identify "step waveform" and "linear waveform",which better reflect the difference of signals.Secondly,for the signal recognition of linear frequency modulation(LFM),Barker,Costas,Frank,P1,P2,P3 and P4 codes,a recognition method based on JTFA and artificial bee colony-support vector machine(ABC-SVM)is proposed.The method adopts features based on JTFA and the classifier is ABC-SVM.It can recognize eight radar signals with fewer features at low SNR.The simulation results show that the method has excellent robustness and less time.When signal-to-noise ratio is-4 dB,the classification rate is 94%.Finally,a recognition method based on AF is proposed.This method adopts partial features,CWD features and features based on the main ridge slice of AF.It is used to recognize twelve kinds of radar signals,that is,LFM,Costas,Barker,Frank and P1-P4 and T1-T4.The classifier is ABC-SVM,which includes two networks.The network 1 divides signals into nine types,that is,LFM,Costas,Barker,T1,T2,T3,T4 and polyphase codes by time features,AF features and partial JTFA features.When the signal is considered to be thepolyphase codes,network 2 classifies them in detail by JTFA features.The experimental results show that the classification rate is 96.2% when the signal-to-noise ratio is-4 dB.