N.N. Lebedeva, O.A. Sidorova, R.A. Maragei, A.N. Kotrovskaya

The problem of recognition of different types of human emotions on the basis of acoustic speech characteristics is of interest both in the theoretical aspect and for solution of different applied tasks. It is especially important for the objective definition of a human state by the tune of his/her voice in different fields of activity, in particular, under extreme conditions when a speaker is beyond the visibility range. The authors of the article developed the software for the analysis of emotionally colored speech signals, which makes it possible to calculate spectral speech characteristics, such as peak frequencies (formants, overtones) and the spectral power density, and to plot the noise approximant of a spectrum for determination of the fractal dimension of speech. The fractal analysis was applied to actors - speech signal in simulated emotional states. Significant differences in fractal dimension were revealed between the simulated states of sorrow, joy, and anger as compared to a neutral state. Moreover, some gender differences were shown in the dynamics of the fractal dimension. In men, the most prominent changes in the parameter under study were observed during simulation of sorrow (a decrease of 30% as compared to the neutral state) and anger (an increase of 18%), whereas simulation of joy virtually did not change the fractal dimension. By contrast, in women, simulation of sorrow decreased the dimension by 16%, simulation of anger was associated with only a 9%-increase, whereas simulation of joy by women increased the fractal dimension of speech by 21% as compared to the neutral state. Taking into account that the growth of the fractal dimension increases "the diversity" within the respective process, i.e., its intonational modulation, which brings the process under study closer to random processes, whereas the fall of this parameter, on the contrary, increases its similarity with determinate (fully defined) processes, it can be suggested that sorrow is associated with speech monotony, and more active emotional states involve an increase in intonation diversity. Analysis of the fractal dimension of a speech signal on the basis of its acoustic characteristics suggests that the emotions directed "outward" (for communication) are accompanied by greater intonation variety, i.e., by an increase in the fractal dimension of emotionally colored speech. Sorrow is a negative emotion directed "inward". It is thought to be opposed to joy, one of the main positive reactions directed "outward", from oneself. Anger, though being a negative reaction, is bright and active and also directed outward (from oneself).