The development of computer vision technology provides more feasibility for the inheritance and development of Guanzhong Drum and Gong Dance, a traditional folk dance. This paper analyzes the dance movements of Guanzhong Drum and Drum Dance, summarizes the artistic characteristics embedded in the dance movements, and assists in the construction of the inheritance path of the dance as the research idea. The basic features of Guanzhong Drum and Drum Dance are briefly analyzed to pave the theoretical framework of the study. Afterwards, the movement characteristics of Guanzhong Drum and Drum Dance are converted by using the universal format BVH, and the pre-processing of the dance movement data is carried out. At the same time, 3D CNNs are used as the action recognition algorithm, and the ReLU function is used as the activation function to prevent the model from overfitting, so as to propose a Guanzhong Gong and Drum Dance action recognition model based on 3D CNNs. Compared with similar model algorithms, the proposed model algorithm can recognize the Guanzhong gong dance with an accuracy of 85.00% and above, which shows a superior performance in action recognition.
