Analysis of speech signal processing and research methods for establishing the psycho-emotional state of a person

The purpose of the research is a comprehensive analysis of the speech signal to identify its parameters most in formative for determining the speaker's psychoemotional state. The work is aimed at the comparative study and demonstration of the capabilities of various signal processing methods (based on autocorrelation function analysis, spectral, wavelet, and attractor analysis) in the context of extracting biomarkers of psychoemotional states. Methods . The work employs a set of methods for time-frequency and nonlinear analysis of the speech signal, implemented in the Matlab environment: autocorrelation analysis – to assess signal periodicity and determine funda mental frequency, the parameters of which (mean value, variance, nature of changes) serve as emotion biomarkers; spectral analysis (spectrogram) – for visualization and analysis of changes in the spectral composition of the signal over time, identifying formants and articulation features associated with emotional coloring; wavelet analysis (using the Morlet wavelet) – for investigating time-localized features of the signal at different frequency scales, allowing the analysis of transient processes and energy distribution across frequency bands; attractor analysis (dynamic systems analysis) – for reconstructing the phase space of the speech signal and investigating its nonlinear dynamics through the construction and analysis of attractors (their shape, complexity, fractal dimension). Results . The effectiveness of each method in revealing specific biomarkers of emotional state was confirmed. It was shown that the combined application of the methods provides a deeper and more complete understanding of the speech signal than each method individually. It was concluded that attractor analysis is the most promising method for the primary detection of profound changes related to psychoemotional state, due to its ability to reveal nonlinear and hidden patterns inaccessible to traditional linear methods. Conclusion . The research demonstrates that modern speech signal analysis methods, from classical ones (autocorrelation function, spectrogram) to advanced ones (wavelet and attractor analysis), provide a powerful toolkit for extracting objective emotion biomarkers. A combined approach allows for a comprehensive assessment of speech dynamics. The potential of attractor analysis is particularly highlighted as a method capable of detecting subtle nonlinear patterns, which opens new opportunities in developing systems for automatic determination of psychoemotional state, diagnosis of disorders, and creating more natural human-machine interfaces