A Human Pose and Attention Mechanism Method for Behavior Recognition and its Application

Behavior recognition is an important research topic in computer vision. Extracting human motion features from videos and recognizing behaviors can provide significant support in fields such as education, healthcare, and security, with broad application prospects. However, in the context of teacher training recognition, mainstream behavior recognition methods currently perform poorly due to complex scenes, numerous interference factors, and the lack of public datasets. To address this issue, this paper proposes a behavior recognition model for teacher training. This model uses a Transformer-based network with an attention mechanism to extract the features of behaviors from teacher training videos. Additionally, it leverages behavior pose information, which is complementary to visual descriptions, to capture high-level behavior information embedded in deep features. To validate the model's effectiveness, a teacher training behavior dataset was constructed, and experiments were conducted. The experimental results indicate that the proposed model exhibits good performance in recognizing teacher training behaviors in higher education institutions.