Time-Frequency Image Processing For Radar Emitter Signals

The classification of radar emitter signals is an important research task in modern electronic reconnaissance and electronic support system, and also determines the technical merits of electronic reconnaissance equipment. Recently, with the countermeasure activities in modern electronic warfare becoming more and more drastic, the classification of the advanced radar emitter signals has become the crucial technique and the difficulty in signal processing of electronic warfare. As a novel modern signal processing technology, time-frequency distribution that transforms time-domain signals into two-dimension time-frequency signals, not only reflects the distribution of emitter signal energy in time and frequency plane, but also reveals the change of emitter signals' frequency with time, so it provides the important information for the feature analysis and the classification of emitter signals. But the time-frequency distribution is very complex for the advanced radar emitter signals and easily disturbed by noise in the process of transmitting and processing. Moreover, the Signal-to-Noise Rate (SNR) of emitter signals is always changeable. So it is very difficult for the processing of radar emitter signals in time-frequency domain.Therefore, considering the processing problem of advanced radar emitter signals time-frequency distribution at low SNR, this thesis discusses the time-frequency distributions of radar emitter signals as a grayscale image, and then processes the time-frequency image of advanced radar emitter signals using image processing technology. The dissertation mainly studies the time-frequency image pre-processing, the time-frequency analysis of advanced radar emitter signals and the classification methods for radar emitter signals based on time-frequency image features. The main work and research fruits are as follows.1. The mathematical model of advanced radar emitter signals is analyzed systematically. Considering several typical radar emitter signals, the Wigner-Ville distribution and Cohen's time-frequency distribution are used for time-frequency analysis of the emitter signals and the corresponding experimental results are given.2. To solve the limitations of traditional image-enhancing methods, a novel method for image enhancement is presented based on rough set theory. Experimental results show that the introduced method can remove noise effectively, and preserve the details of image edges to a certain degree and improve the contrast. This method is superior to traditional methods in terms of image enhancement effects and time complexity. And then this image enhancement method is adopted to enhance the time-frequency image of the advanced radar emitter signals and the experimental results are satisfying.3. To overcome the limitations of conventional time-frequency analysis methods for processing multi-component radar emitter signals, a novel method for time-frequency images processing of multi-component radar emitter signals is proposed based on image processing technology in this dissertation. Based on analyzing the time-frequency distribution of multi-component radar emitter signals, the method regards the time-frequency distribution of emitter signals as grayscale images, and makes use of the spatial smoothing operation, the threshold comparison and the thinning algorithm of morphology. Simulation results show that the introduced method can remove noise effectively, and improve the time-frequency image's resolution. The method is quite suitable for analyzing the multi-component radar emitter signals in time-frequency domain and it is superior to conventional time-frequency analysis and the time-frequency reassignment methods.4. Using neural network to process time-frequency image and extracting the modulation parameters of emitter signals in the time-frequency image, this dissertation presents a method for time-frequency image processing of multi-component advanced radar emitter signals. Simulation results demonstrate that the method can effectively process the time-frequency image and can exactly extract modulation parameters of multi-component emitter signals at low SNR.5. To correctly classify advanced radar emitter signals, this dissertation analyzes time-frequency image features and presents a novel method for classifying radar emitter signals based on time-frequency image features. This method transforms the classification of emitter signals into image processing and image recognition. Emitter signals are analyzed in time-frequency domain and their time-frequency images are obtained. These images are transformed into grayscale images and normalized. Support vector machines are applied to design classifiers for recognizing the processed images, which corresponds to different radar emitter signals. Experiments conducted on five typical emitter signals show that the introduced method achieves more than 92% correct classification rate on condition that the signal-to-noise ratio is above 2.5dB.