A Behavior Recognition Algorithm Based on Global Context Awareness and Pose Information and Its Application in University

Behavior recognition is an important technology for computer systems to recognize and understand human activities. As an important research topic in computer vision, it has significant application value in areas such as education, human-computer interaction, medical rehabilitation, and autonomous driving. However, in examination scenarios, the complexity of the scene, numerous interference factors, and lack of public datasets lead to poor performance in current mainstream behavior recognition methods. To address this, this paper proposed a behavior recognition model for university students in examination settings based on global context awareness and behavior pose information. This model considers the issue of diminishing examination behavior features with increasing network depth and leverages the advantage of Transformers in capturing long-range dependencies to design a global context information module. It uses behavior pose information that complements visual descriptions to capture high-level behavior information from deep features. To validate the model's effectiveness, an examination behavior dataset was constructed, and experiments were conducted. Experimental results show that the proposed model demonstrates superior performance.