Bielozorova Y. Method of application of wavelet analysis in problems of identification of speech information

The thesis is devoted to the development of the model and method that process and determine the individual characteristics of the person and the identification of the speech signal. Models and methods are based on the apparatus of fractal and wavelet analysis. The paper substantiates the need to use fractal and wavelet analysis to identify speech information. The methodology using fractal signal characteristics has been developed for speech signal segmentation. The application of the methodology allows to distinguish vocalized and unvocalized fragments of the signal. This happens regardless of the person's language data. The algorithm for selecting parameters of selfsimilar structures in a speech signal has been proposed, it based on wavelet transform maxima at different levels of voice signal decomposition. This allows us to determine the frequency parameters of the fundamental tone and the formant envelop in the form of probability density curves. The method of personal identification has been developed. This method uses the methodology of speech signal segmentation and the algorithm for selecting parameters of self-similar structures. The software system was created on the basis of the performed research. This software system automatically calculates the voice characteristics of the voice records. In addition, it performs the ranking of these characteristics in the database, according to the criteria, which was defined in the work and the identification of the person in the speech signal.