A Music Playback Algorithm Based on Residual-Inception Blocks for Music Emotion Classification and Physiological...

Area of Science:

• Exercise Physiology
• Computational Intelligence
• Music Information Retrieval

Background:

• Music is known to enhance athletic performance and motivation.
• Practical applications of music interventions in exercise are underrepresented in research.
• Existing systems lack integration of real-time physiological data and emotional music analysis.

Purpose of the Study:

• To design an energy-efficient music playback system for joggers.
• To integrate artificial intelligence (AI) with physiological signals and emotional music for personalized selection.
• To improve the practical implementation of music during exercise for enhanced user efficiency.

Main Methods:

• Developed a music selection module using AI techniques, physiological data, and emotional music analysis.
• Optimized the model using logarithm scaled Mel-spectrograms for reduced computational complexity (FLOPs) and parameters.
• Evaluated model performance on Bi-modal, 4Q emotion, and Soundtrack datasets.

Main Results:

• The proposed model achieved competitive accuracy: 84.91% (Bi-modal), 92.04% (4Q emotion), and 87.24% (Soundtrack).
• Demonstrated significant reductions in computational complexity and inference time compared to existing models.
• The model's small size allows application on mobile devices with limited resources.

Conclusions:

• The designed playback system effectively links music emotion with physiological states during exercise.
• The system offers a practical solution to improve users' exercise efficiency and motivation.
• The AI-driven, low-power approach is suitable for real-time application on portable devices.