The goal of this work was to study the possibility of use of speech biometry for identification of speaker, determination of speech accent, and estimation of the emotional state of the speaker. It was suggested to solve these problems by analysis of the two main components of the linear model of speech production: pitch and formant frequencies. It was shown that pitch and formant analysis provided estimation of the emotional state of the speaker. It was found that sthenic emotions were accompanied by an increase in the pitch and its dispersion. The speech accent was determined using formant analysis and neural network classifiers. Speaker identification was successfully performed in a test group of 50 subjects.