The article describes the method of solving the problem of identifying a person's voice. We consider the methods of analysis and the audio coefficient systems. Algorithms for signal preprocessing and selection criteria are described. Network oncoming distribution and Kohonen maps modifications.