Morozov

The work deals with the problem of identification by voice. The focus is on building an adaptive classifier of specific voice features based on the Kohonen self-organizing map. The authors present the algorithms used for audio processing and classification, as well as the software system VOICE which implements these algorithms. The results of the studies confirming the effectiveness of the algorithmic and software solutions are given in this paper.

Dynamic system attainability domain approximate construction problem based on “multi-finish” and model differential equation system vector field approximation methods combination is discussed. The main point of the first method is multiple model differential equation system integration with different controls values. The idea of the second method is preliminary approximation  of model differential equation system’s right side. The purpose of this article is to find out the efficiency of neural network approximation. The research is carried out for 7^th order differential equation system, which describes aircraft center-of-mass motion. Three-layer feed-forward neural network and similar cascaded neural network are analyzed.