Algorithm for Classroom Behavior Recognition in Learning Spaces with Technological Support

The study aims to improve teachers' teaching outcomes by integrating teachers' teaching methods and promoting teachers' teaching activities of different ages in learning spaces with environmental advantages supported by technology, and changing the emotional interaction modes and standard behavior patterns of teachers through the transformation of learning spaces supported by technology. Based on the idea of double-flow method, the study first obtains the skeleton keypoint feature data of the student target using the AlphaPose human pose detection model. Then, the student target's RGB image cropping image is obtained through adjusting the non-maximum suppression algorithm of the YOLOv3 model, and the feature data in the cropping image is extracted using the ResNet18 network with added attention mechanism. Finally, the two feature data are fused and input into the fully connected layer for classification, thereby constructing a student behavior and teacher behavior recognition model based on data fusion.