S.Y. Chelyshov, A.V. Konichenko, A.S. Sizov, A.O. Bogomolov

At present, special importance of ensuring continuous monitoring and reliable «alarm» from quickly deployment systems of unauthorized appearance detection in the protected zone of the specified classes objects. To ensure the high secrecy of the functioning, the reduced deployment time, the large controlled area allow sensors-classifiers on the basis of seismic sensors. On the basis of the conducted research the authors developed a method of class recognition of moving objects with the correlated frequencies spectra. The method is based on a two-stage procedure of sequential processing vector space of attributes that describe the spectrum of the signal, and the signs, describing the frequency-time structure of the signal. However, at the first stage of the procedure applies the criterion of minimum Euclidean distance with weights corresponding values of the variances of characteristics, but on the second recognition is carried out on the basis of the analysis of the rules of the impact that connect the values of attributes and recognition object classes. In the framework of the stage of preparation of the initial data for the decision of problems of recognition consistently defined: the statistical model of object classes on the basis of the information components amplitudes of the frequency spectrum; the matrix of frequency of occurrence of pairs classes, located at a minimum distance from recognized implementations, belonging to the learning sample of objects of the i-th class; the model of object classes, built in space of the variances of the wavelet coefficients. In the result of the research proved the expediency of application of the first stage of the procedure of recognition of the object's class - weighted Euclidean distance between the implementation and the standards of class objects. At this stage there are two classes of objects, with the lowest value of the weighted Euclidean distance to the obtained implementation. At the second stage for recognition of objects with similar frequency spectra apply obtained at the stage of preparation of the initial data model, in which the attributes are applied dispersion of wavelet coefficients. The criterion for recognition is the highest value of the coefficient of confidence. Thus, the proposed method allows to solve the problem of recognition of land of moving objects. The capabilities of existing means of digital processing of signals (microprocessors and signal processors) allow to realize the developed procedure for the recognition of the Autonomous sensors-qualifiers.