Research On Identification Of The Radar Signal Based On Time-Frequency Analysis

With the continuous development of radar technology, various radar signals havebecome more complex. The new complex radar emitter signal has been dominant andthe traditional five parameters have been unable to meet the identification of radarsignals. Coupled with more intensive electromagnetic signal density environment, moreeffective identification features is needed.Time-frequency analysis is a new and effective method of non-stationary signalprocessing, from two dimensions to reflect the time domain and frequency domainsignal energy changes. In this paper, for complex modulated radar signals,time-frequency analysis is used to identify radiation signal by extracting thetime-frequency entropy characteristic and instantaneous frequency characteristics.Among the many time-frequency analysis methods, a new kernel distribution withbetter performance of time-frequency anlysis methods is used to analysis the signal.Then entropy feature of time-frequency distribution is extracted as the signalrecognition features. The effective identification of the typical six kinds of radar signalindicated the effectiveness of entropy characteristics. Meanwhile, the joint applicationof the kernel distribution analytical method, the average recognition rate is icreased by2%.On the base of using the peak of time-frequency distribution based IF estimationmethod, IF cascading feature has been extracted to constitute a feature vector, as a radarsignal recognition feature. After optimization of the selection and arrangement of thefeature parameters, the hierarchical tree was rebuilt and different modulation radarsignals have been effectively identified. Joint application of time-frequency distributionbased IF estimation method and the rebulit hierarchical tree constituted of the featurevector has improved the lowest radar signal recognition rate of5%under the conditionsof3~21dB. Using the paper proposed two-stage IF sequence difference characteristicsbetween the center point value and the mean vaule of the beginning point and the endpoint, LFM and one class NLFM such as SFM whose IF sequence is centrosymmetrichave been effectively identified.