Intelligent attendance monitoring system with spatio-temporal human action recognition

This paper proposes an intelligent attendance monitoring system based on spatio-temporal human action recognition, which includes human skeleton gait recognition, multi-action body silhouette recognition and face recognition. Our system solves several problems, for example, when a mask is worn to conceal the face, which leads to a decrease in recognition accuracy performance, and when a 3D face mask is used to fake an identity. The skeleton gait feature of our intelligent attendance monitoring system uses a temporal weighted K-nearest neighbours algorithm to train the recognition model and carry out identification, while the multi-action body silhouette feature uses a multiple K-nearest neighbours algorithm to train the recognition model, identify the person and vote on the outcome. Using the proposed system, which integrates skeleton gait features, action silhouette features and face features, more effective recognition can be achieved. When the system encounters a situation with feature masking, such as when an individual is wearing a mask or has changed their clothes, or when the viewing angle is masked, it can continue to deliver good recognition ability through multi-angle skeleton synthesis gait recognition. Our experimental results show that the recognition accuracy of the system is 83.33% when a specific person wears a mask and passes through a monitored area. The intelligent attendance monitoring system uses a LINE messaging API as the access control notification function and provides a responsive web platform that allows managers to perform follow-up management and monitoring.