Research on interactive optimization technology in music education games based on EMD-RNN

Current music education games are challenging in meeting students' personalized learning needs. Interactive optimization technology based on EMD-RNN provides new possibilities for solving this problem. Based on the EMD-RNN model, this study explores its application of interactive optimization technology in music educational games. The EMD (Empirical Mode Decomposition) method adaptively decomposes audio signals into multiple eigenmode functions, which makes the processing of non-stationary signals more accurate, especially suitable for intonation and rhythm analysis in music teaching. Combining the advantages of EMD and RNN, this study proposes an innovative interactive optimization framework that dynamically adjusts game difficulty and learning tasks by analyzing students' audio performance, interactive data, and learning progress in real-time. There are apparent differences in students' performance at different stages. In the rhythm training task, the average performance of students is 57.2%, and some students have significant intonation errors, resulting in large fluctuations in scores. 91.6% of the students performed well in the note recognition task, while 23.7% performed less than satisfactorily in the same task, showing a significant difference in intonation training. This framework not only provides personalized teaching feedback to optimize students’ learning paths based on their rhythm deviation and intonation accuracy but also significantly enhances student engagement and motivation through adaptive challenge adjustments. Experimental results demonstrate that the proposed method achieves a 4.6% error rate in note recognition and a 63.4% intonation correctness rate, outperforming traditional models in personalized music education.