S.I. Andronov, O.V. Rodionov, I.V. Chebotar

The article is about an algorithm of separation of groups of structurally-similar signals. Application of the combined system with the self-training, constructed on the basis of ideology of the self-organizing classifiers is offered in this article. Algorithm testing was conducted by a method of imitating modelling. To solve the formulated problem specialized software has been worked out by «LABVIEW - 8.0» imitating modeling. The probability of an error of separation of structurally-similar signals was thus estimated. The criterion of efficiency of the given procedure the probability of correct separation had been accepted. The worked out algorithm allows to lower size of an error of decision-making, to raise probability of correct separation of elements of a group signal and to provide faster convergence to algorithms of training on learning samples of structurally- similar signals.