With the increasing maturity of deep learning technology, human action recognition based on deep learning has received extensive attention from research scholars. In this paper, based on convolutional neural network and biomechanics theory, the recognition and characterization of musical instrument playing actions are studied. In terms of action recognition, this paper improves the GoogLeNet network structure and constructs a musical instrument playing action recognition model. On this basis, human biomechanical modeling research is carried out. The results show that the average recognition rate of the method proposed in this paper on the publicly available image dataset PPMI is relatively high, reaching 66%, which is better than other comparative methods, confirming the feasibility and effectiveness of the model application system for human action classification. The results of biomechanical modeling analysis show that the reasonable allocation of the time ratio affects the basic rhythm of the movement, and the adjustment of the rhythm of the center of gravity displacement and center of gravity velocity not only affects the basic rhythm of the whole musical instrument playing movement and the quality of the movement, but also the basic rhythm of the movement and the requirements of the movement constrains the allocation ratio of the center of gravity displacement and center of gravity velocity at each stage.
