Optimization of music education strategy guided by the temporal-difference reinforcement learning algorithm

To make up for the shortcomings of traditional music teaching strategies and improve the intelligence of music teaching, this study uses a reinforcement learning (RL) algorithm to conduct an intelligent exploration of Erhu teaching methods in music. Firstly, a rule-based Erhu fingering evaluation method is proposed, which summarizes the fingering habits and general rules of modern Erhu performance and constructs a quantitative evaluation system (QES) of Erhu fingering. This system provides the evaluation basis for effectively verifying the intelligent generation model of Erhu fingering proposed here. Secondly, on the one hand, an intelligent generation model of Erhu music is proposed based on neural network technology. On the other hand, an intelligent automatic generation (AG) algorithm for Erhu fingering is put forward. In this algorithm, the temporal-difference RL (TDRL) model and off-policy are integrated, and the influence of the fingers before and after actual playing is considered comprehensively. Finally, the validity and feasibility of the proposed Erhu music generation model and the Erhu fingering-intelligence generation model are verified by simulation experiments. The results reveal that: (1) The QES of Erhu fingering proposed here can objectively describe the advantages and disadvantages of Erhu fingering and play a role of feedback and improvement to the generation model of fingering; (2) In the proposed Erhu music generation model, the musical note index value of the generated music is high, which avoids the situation of excessive note repetition and note jump amplitude in the generated music. (3) The designed Erhu fingering-intelligence generation model is employed to compare and analyze three kinds of music segments. It is found that the total score and scoring rates of fingering evaluation generated by the three pieces of music are relatively high and very close to the professional fingering, scoring rate difference is less than 3%; (4) The scoring rate of all kinds of fingering generated by machines is about 90%, and the difference with professional fingering is no more than 3%. The data show that the proposed method can realize the AG of Erhu fingering well. This study aims to assist in music and fingering teaching for Erhu course education and offer some references for other courses in music teaching. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2024.