Research On Feature Extraction Of Micro Motion Target And Reconstruction Of Incomplete Signal

When radar illuminates a moving target,if the target or the components still has micro motion like vibration or rotation,then the micro motion will produce Doppler modulation of echo,that is called micro Doppler effect.In recent years,micro Doppler effect is widely used in radar target classification and recognition.In this dissertation,we thoroughly study the feature extraction methods based on the characteristics of the micro Doppler in order to get more effective features.According to the actual radar dwell time is short,it is difficult to obtain the continuous observation on the same target for long time,namely the problem of incomplete signal.Inspired by that problem,we improve the signal reconstruction algorithm in order to improve the classification performance.The main contents of the article can be summarized as the following three parts:The first part,the radar echo parameter model of the rotor is introduced according to the existing research results,then time domain and Doppler domain echoes of helicopter,propeller aircraft and jet aircraft are simulated,and the difference of micro Doppler characteristics of three kinds of aircraft is analyzed.In addition,according to the characteristics of pedestrian and vehicle target,the pedestrian and vehicle targets are modeled,and the differences of the micro Doppler characteristics of the vehicle and the vehicle are analyzed according to the measured data.The second part,the basic concept of fractional Fourier transform is introduced,and it is introduced into feature extraction of micro motion signal.Then,two methods are used to select effective features and remove redundant features.The experiments of the simulated data and the measured data show that the fractional domain features classification has better performance compared with the traditional characteristics of time domain and Doppler domain,and the requirements of the pulse repetition frequency,dwell time and signal-to-noise ratio(SNR)are relatively low.The third part,in view of the fact that there are some defects in the narrow band radar signal acquisition,which leads to the decline of the classification performance,various signal reconstruction methods are studied.Then,a new reconstruction method based on complex Gaussian model is improved,which can reconstruct the signal time-frequency spectrum directly,which is convenient for the subsequent reconstruction performance verification and time-frequency feature extraction.On this basis,the noise robust model is improved,so that it can be reconstructed under low SNR.Finally,the experimental results based on the measured data show that the proposed method can achieve good reconstruction performance,and obtain the correct classification of human and vehicle data in noisy condition.