Prodeus A. Theoretical substantiation and practical implementation of digital sonar signal classification systems

Dissertation is devoted to creation of theory basics of computation and designing of digital sonar signals classification systems. The expediency of use of expert systems technology both at a stage of designing, and at a stage of decision making, is shown in the dissertation. Useful for engineering applications computational relations are obtained as a result of analysis of methodical errors which are inherent to measurements of classification parameters by threshold and approximative methods. Coefficient of excess of a harmonic or linear frequency modulated echo signals is offered as sonar classification parameter. It is shown that the error of a traditional estimate of fourth moment of echo signal depends on a position of an echo signal spectrum on frequency axis and may be unacceptable high. Two new estimates of fourth moment of echo signal, which are free from the mentioned defects, are offered. Decision making rule, where sets of objective and subjective signs are used and which is based on usage of sequential Bayesian approach, is offered and investigated. The approaches to computation and designing of sonar classification systems, which essentially use experience of designing and exploitation of the working classification systems, are offered. Special kinds of expert systems which are destined for developing and optimization of digital sonar classification systems software are developed. Keywords: classification signs, decision making, digital sonar, expert system, functional software, parameters measurements, signals classification.