Algorithm for Dynamic Fatigue Detection of PC Operator Based on Gaze Characteristics

Abstract

The onset of fatigue is dangerous in areas of activity that require high concentration of human attention, such as air traffic controllers, nuclear power plant operators, etc. It should be noted that these types of activities are characterized by the fact that most of the time the employee sits at the workplace and his gaze is directed at the monitor. The paper presents the algorithm for dynamic fatigue detection of PC operator based on eye movement characteristics. The algorithm for dynamic fatigue detection includes preparation stage and fatigue detection stage. Within the preparation stage, gaze movement characteristics are calculated and correlations with fatigue test results are searched for. Within the fatigue detection stage, eye movement characteristics that most strongly correlate with fatigue are selected. These characteristics can also be divided by the types of physical events on which they are based. We can distinguish such characteristics as speed, percentage, saccade length, curvature of gaze trajectory. To find correlations between eye movement characteristics and fatigue, a dataset of eye movements and the results of tests and questionnaires such as the go/no-go task, the “Landolt rings” test, and the VAS-F questionnaire were analyzed. The dataset consists of gaze coordinate recordings from 15 participants acting as PC operators. To assess the degree of fatigue, the participant completed the VAS-F questionnaire. The “Landolt rings” test is a test used to measure attention concentration. The labeled dataset is used to train a machine learning model that detects fatigue. The experimental results showed that using the characteristics selected in the study yielded the most promising results. This approach allowed us to achieve the highest F-score and the best average accuracy, indicating the overall reliability of the model.

Keywords

• fatigue detection
• artificial intelligence