An artificial intelligence-based classifier for musical emotion expression in media education

Area of Science:

• Artificial Intelligence
• Music Psychology
• Educational Technology

Background:

• Music effectively conveys emotions and regulates mood, crucial for learning.
• Current AI music emotion analysis often relies on lyrics, neglecting audio signal processing.
• Limitations exist in AI's perception, transmission, and recognition of music signals for emotional analysis.

Purpose of the Study:

• To develop an AI-based music emotion analysis method using audio features.
• To improve teaching efficiency through systematic emotional assessment in music education.
• To address limitations of lyric-dependent AI models in music emotion recognition.

Main Methods:

• Intelligent segmentation and note recognition using sound-level processing and threshold determination.
• Construction of a music emotion classifier utilizing a Radial Basis Function (RBF) model.
• Training the classifier with correlation feedback and comparing with the Hevner emotion model.

Main Results:

• Accurate recognition of audio features in 0.004 min.
• Achieved an accuracy rate exceeding 95% for music emotion classification.
• Classified emotions into four categories: Quiet, Happy, Sad, and Excited.

Conclusions:

• Audio feature-based emotion classification aligns with human cognition.
• The developed AI model offers a more robust approach to music emotion analysis.
• This method has potential to enhance music education through improved emotional understanding.