Research on multi-label classification model design in online teaching and learning for music scholars from a biomechanical perspective

Music serves as a vital medium for emotional expression and cultural heritage, evolving significantly through advancements in digital education. This study introduces an integrated framework to enhance online music education via innovative music generation and genre classification techniques. Central to this research is the MGED (Music Generation and Education Development) framework, which utilizes Convolutional Neural Networks (CNNs) for feature extraction and Highway Networks for deep learning. By incorporating spatial attention mechanisms with Bidirectional Gated Recurrent Units (Bi-GRU), the framework generates high-fidelity spectrograms, while the Griffin-Lim algorithm ensures temporal coherence in the outputs. For genre classification, the study employs the CMBRU (Classification Model for Bi-GRU and Residual Units) framework, leveraging Mel-Frequency Cepstral Coefficients (MFCC) and multi-channel CNNs to achieve robust representation learning. This model effectively captures temporal dependencies, resulting in over 70% accuracy across five genres. Additionally, this research explores the design of a multi-label classification model tailored for online teaching and learning environments aimed at music scholars, viewed through a biomechanical lens. As online education becomes increasingly prevalent, the need for effective classification models that can handle multiple labels simultaneously is critical, particularly in music education, where diverse skills and knowledge areas intersect. The study employs biomechanical principles to analyze the physical aspects of music performance and learning, integrating these insights into the classification model. This approach not only enhances the educational experience for music scholars but also contributes to the broader field of online music education, paving the way for future research and practical applications.