Дослідження алгоритмів розпізнавання тендерної приналежності людини за її голосом

Abstract

Object of research - methods of gender recognition using speech signals. The purpose of this work is a research and implementation of methods for automatic gender identification by voice. The main methods of gender recognition using speech signals are: method of support vector machines, method of gaussian mixture model. A comparative analysis of methods has been made. Voice features and a method of their extraction on a basis of mel-frequency cepstral coefficients were reviewed. Also algorithms of gender recognition through speech were reviewed. In the practical part the algorithm of gender recognition was implemented based on the frequency of the main tone of voice.