Oscillatory-Neuron-Enhanced YOLO for Robust Object Detection and Its Application to Fine-Grained Classroom Behavior Recognition

Classroom behavior recognition, as a key component of intelligent monitoring in educational informatization, demands high levels of real-time performance, accuracy, and robustness. However, traditional YOLO-based object detection algorithms tend to suffer from performance degradation in complex classroom environments due to variations in illumination, occlusion, dense target distributions, and fine-grained action ambiguities. To address these challenges, this study proposes an enhanced YOLOv9e framework by embedding five types of oscillatory neuron modules—CosineFocus, AdaptiveThreshold, SineEnhance, DampedResonance, and BurstPulse. By integrating the dynamic modulation mechanisms of oscillatory neurons into the feature flows of both the Backbone and Neck, the proposed approach strengthens feature selectivity, noise suppression capabilities, and local sensitivity. File Description:Classroom-Behavior-Detector(source code).rar is the source code. Student Classroom Behavior1.rar and Student Classroom Behavior2.rar are datasets of Student Classroom Behavior; Detecting Student Classroom Behavior with Oscillatory.rar contains a dataset of student classroom behaviors used to demonstrate the effectiveness of the oscillatory neuron network. Guoming Chen and Shijin Lin contributed equally to this work and are co–first authors.