Application of gesture recognition based on domain transfer learning and SimSiam network in education management

As one of the important components of augmented reality system, gesture recognition is widely used in virtual reality, intelligent interaction and human-machine interface. In education management, how to improve the level of teaching management and teaching quality, gesture recognition plays a very important role. In view of the problems of negative transfer and poor model generalization in single-source domain transfer learning, which have great impact on gesture recognition, this paper proposes a new novel gesture recognition method based on domain transfer learning and SimSiam network. Firstly, the gesture data set is passed into the SimSiam self-supervised network for training. On this basis, the technique of domain specific feature alignment and domain classifier alignment is adopted. This approach aims to enhance the model’s gesture recognition performance between different users, thus significantly improving the accuracy of cross-user gesture recognition systems. Experimental results show that this proposed method can effectively identify a variety of dynamic gestures with Angle deflection. Compared with the traditional displacement feature method, the average accuracy of the proposed method is increased by 4%, and it can effectively deal with the dynamic gesture recognition problem in the case of palm deflection.