An investigation on the application of deep reinforcement learning in piano playing technique training

The article applies the knowledge of the physiological structure of finger movement in piano playing and the DH parameter method to construct a kinematic model of finger keystroke in piano playing. The Leap Motion-based gesture recognition algorithm is used for piano playing gesture movement extraction, gesture movement sub-framing, and velocity direction encoding, and the judgment HMM algorithm and the Viterbi improvement algorithm based on the a priori knowledge of fingering are utilized to automatically standardize piano fingering. After comparing the recognition performance of the piano playing training model in this paper, it is discussed how it can be used and how it can improve the user’s piano playing skills. In comparison to other piano gesture recognition models, the recognition accuracy and gesture dynamic information description of this paper’s training model demonstrate optimal performance. After the actual playing training experiments, the experimental group achieved significant improvements, while the control group experienced negligible improvements. After the experiment, the difference between the two groups increased, and the experimental group was superior to the control group in all aspects of piano playing. The training model for piano playing in this paper is effective in enhancing the user’s piano playing level.