Recognition of human fatigue based on speech analysis using neural network technologies

Qualitative psychophysiological research studies are associated with the creation of accessible and well-organized databases that require a lot of preliminary work on the development of measuring complexes, including not only tools for measuring the psychophysiological parameters of a human, but also their emotional state, which is displayed in facial expression, speech and behavioral patterns. Measuring systems should also include the means of experimental material processing. The purpose of the study was to conduct an experiment on creating a prototype of the Speech Data Base of Russian-speaking respondents and to obtain answers to some methodological questions that arise among specialists when they use the database for the task of recognizing the state of human fatigue. A hardware and software complex has been developed that allows to synchronously register psychophysiological parameters, video recordings of behavioral reactions and audio recordings of human speech. As a model of physical fatigue, a cardio-respiratory test with physical activity (load) was used. Before and after completing the test, volunteers read out a set of standard phonetically representative texts. The obtained audio recordings were processed using a specialized neural network capable of analyzing the integral spectral characteristics of sound. The results of the experiment showed the possibility of recognizing the state of human fatigue based on speech analysis, which makes it possible to proceed to the creation of a large bank of audio recordings and the improvement of algorithms for recognizing the state of fatigue