Abstract: It is difficult to achieve real-time and relatively accurate monitoring tasks by using GAN or 3DCNN in the actual monitoring edge devices. An abnormal behavior recognition algorithm based on knowledge distillation in examination room is proposed. Compared with the abnormal behavior recognition which takes extracting spatial and temporal features and fuses as the mainstream idea, the abnormal behavior recognition method using video frames for target detection and knowledge distillation is faster and more accurate. With the help of knowledge distillation strategy, the algorithm used pre-trained teacher network to supervise student network learning, infer normal behavior and detect abnormal behavior. The results show that the algorithm reaches the upper level of the mainstream data set, and has good efficiency and accuracy in the examination room environment.