Research On Signal Recognition Technology Based On Time-frequency Image

Due to the interference of signals such as communication signals,radio signals,remote telemetry signals,radar signals,etc.,the electromagnetic signal environment of the battlefield is complex and variable.In order to monitor the electromagnetic signals of the battlefield,it is particularly important to accurately identify the modulation type of the signal from many sources of interference,which is a key step in signal analysis.The signal can be subsequently analyzed only by first identifying the modulation type of the signal,so as to facilitate signal demodulation and other information acquisition.Therefore,in the complex electromagnetic environment,it is of great significance for the identification and detection of signals.This paper focuses on six commonly used radar emitter signals,and identifies the signals from the invariant moments and texture features of the image.At the same time,the LFM signals are detected and estimated,including the following work:(1)Using the characteristics of invariant moments,a wavelet invariant moment feature vector extraction and recognition classification method based on time-frequency distribution is proposed.Firstly,the time-frequency image of radar radiation source signal is processed.The correlation property of wavelet invariant moment is used to extract the eigenvector of wavelet invariant moment from the time-frequency image of radar radiation source signal.Due to the characteristics of a large number of wavelet invariant moments The value is invalid,and the representative eigenvalues need to be selected from it.Finally,the eigenvalues are trained by the support vector machine classification and recognition method to achieve the purpose of signal classification and recognition.The six common radar emitter signals are classified.The simulation results show that the recognition accuracy is still over 90% when SNR=-3d B.Compared with other eigenvalue classification methods based on image invariant moments,this method has better classification effect when the signal-to-noise ratio is lower than-4d B.(2)A radar emitter signal identification method based on Tamura texture feature and support vector machine is proposed.The algorithm firstly performs image preprocessing on the time-frequency image of the signal,and then uses the Tamura texture feature method to extract the features of the processed time-frequency image,and obtains five parameters such as roughness,linearity,directionality,contrast and regularity.The eigenvalues are finally identified and classified by the support vector machine.The simulation results show that the proposed method has better classification under low SNR conditions.(3)Aiming at the parameter estimation of chirp signal with low SNR,a signal parameter extraction method based on Choi-Williams time-frequency distribution image and Hough transform is proposed.Firstly,the time-frequency transform is performed on the chirp signal by CW time-frequency distribution,and the time-frequency image is obtained.The appropriate threshold is selected according to the maximum inter-class variance method for binarization,and then the image is smoothed by morphological method.The image uses the Hough transform,and finally the signal is evaluated based on the result of the peak detection,thereby realizing the estimation of the LFM signal parameters.The simulation results show that the initial frequency estimation error of the single component LFM signal is still less than 0.8% and the estimated value of the frequency modulation slope is more accurate when the SNR is <-6d B.