Neurobayesian Algorithm for Subject's Psychophysiological State Identification

Abstract

At the present stage of the technology development, the reliability indicators of technical systems have increased, while the person reliability began to recede over time, therefore increasing the role of the subjective factor in the emergence of industrial accidents and incidents. In order to reduce the risk of damage made by the subject, the admission to the performance of professional tasks should be multistage and periodic (continuous). Accidents can be the consequence of subjects staying in “inadequate” psychophysiological state: alcohol intoxication, stress, drowsiness, etc. In this work, the authors offer a neuro-Bayesian algorithm for recognizing psychophysiological states of subjects using facial thermographic images, based on the use of convolutional neural networks committees and sequential application of the Bayesian hypothesis formula. More than 97% of the recognition accuracy of seven psychophysiological states has been achieved for a 10 second monitoring period, which exceeds the known world indicators both in accuracy and recognition duration.