A non-contact mental fatigue detection method for space medical experiment using multi-feature fusion model

Abstract

Mental fatigue detection in space medical experiment is an essential and indispensable concerning work efficiency and human security. The non-contact mental fatigue detection methods based on face video become increasingly prevalent due to its low cost and non-intrusiveness. However, the current methods are limited by feature selection and complex model. To detect mental fatigue high efficiently without interruption and uncomfortable feelings caused by contact to test equipment, a novel non-contact mental fatigue detection model using Deep Convolutional Neural Network (DCNN) is proposed in this study, which combines facial feature extraction and multi-feature fusion. The identified facial features are eye aspect ratio (EAR), percentage of eye closure over a period of time (PERCLOS) and blink rate, then multi-feature fusion of facial features, heart rate and global features extracted from facial video was employed for mental fatigue detection. The proposed method was evaluated in the Chinese “Earth -Star II” 90-day HDBR experiment and was compared to state-of-the-art techniques. The experimental results verify the effectiveness and validity of our method.