Application Of Wavelet To Electronic Reconnaissance

Signal processing techniques play an important role in modern electronic warfare, such as parameter estimation of signals, modulation recognition of communication signals, specific emitter identification, and so on.The aim of the dissertation is to research the techniques based on the concept of wavelet for electronic reconnaissance. The dissertation can be concluded as follows:Wavelet can tell us how weighted averages of certain other functions vary from one averaging period to the next. This interpretation of wavelet analysis is a key concept, so we explore it in some depth in the Chapter 2. The use of wavelet transform for modulation identification of digital signals is described. The wavelet transform can effectively extract the transient characteristics in a digital communication signal. So by using the binomial distribution to model the number of signal transient, we can derive the required test statistics and develop the maximum likelihood modulation classification algorithm. A new efficient parallel classifier based on signal instantaneous parameters which can be estimated by using the Haar wavelet transform is also presented.The analysis method can also be used in radar signal parameters estimation.Modulation classification algorithms based on the likelihood functional are discussed in Chapter 3. In general, the problem of classifying between K possible modulations can be modeled as the multiple hypothesis testing problem. The sufficient statistics for modulation classification can be extracted from the detection theory or via the approximated phase/amplitude probability density function. We follow this idea and investigate both the methods. Then a new modulation classification algorithm based on maximum likelihood criterion is introduced for the problem of identifying modulation types in the wavelet transform domain. This classifier is based on modeling the wavelet coefficients by a generalized Gaussian distribution. Numerical experiments are used to illustrate the effectiveness of the proposed method.In Chapter 4, we study the use of wavelet transform to estimate the symbol rate of an M-ary phase shift keying (MPSK) signal. Previous work uses the wavelet transform to locate the transients produced from phase changes. The separationbetween transients gives a symbol rate estimate. Then another estimation method based on Haar wavelet ridges which can detect the signal details effectively is presented. Finally, we develop a multiband or wavelet approach to estimate the symbol rate of MPSK signals. The technique utilizes the simple and elegant nonlinear energy operator to extract the instantaneous amplitudes and frequencies. The method first filters the signal through a bank of wavelet-like passband filters to improve the performance of energy operator which will be used to the largest filter output response to extract the energy transient information in order to estimate the symbol rate. The approach is robust against noise and simple in computation. The theoretical predictions and the simulation results indicate that improved practical strategies are feasible. ' 'The problem of estimating the parameters of complex signals having constant amplitude and polynomial phase, measured in additive noise, is discussed in Chapter 5. Cramer-Rao lower bound for wavelet transform-based polynomial phase coefficient estimates is derived, and a new parameter estimation algorithm is also presented. We also show that wavelet transform based approach can reach much lower Cramer-Rao bound as compared with the existing techniques.Finally in Chapter 6, we give an overview of the radar emitter identification methods. The fuzzy ARTMAP neural network is used to classify streams of pulses according to radar type using their type-specific and position-specific functional parameters. Performance is improved by using such a What-and-Where fusion strategy. We also provide a coding length based on the Minimum Description Length criterion for specfic emitter identification (SEI). A novel mutual information feature extraction approach is proposed in the last section. The set of features are generated from dilations and shiftings of a family of one or more mother wavelet functions. Then the measure of mutual information between the pulse category and feature vector element values over all pulse realizations is obtained based on Parzen window density estimation. The mutual information can be used to discriminate one category from another. The approach has a good characteristic of easy implementation and short computation time.