Modeling emotional content of music using system identification

Area of Science:

• Computational musicology
• Affective computing
• Signal processing

Background:

• Modeling the emotional impact of music is complex.
• Previous methods often lack generalization across genres.
• Quantifying emotion in music requires robust analytical tools.

Purpose of the Study:

• To develop a novel methodology for modeling music's emotional content over time.
• To utilize system-identification techniques for creating predictive emotion models.
• To generalize emotional content modeling for entire music genres.

Main Methods:

• Quantified musical emotion using valence and arousal dimensions.
• Applied system-identification techniques to build time-series models.
• Validated model performance using statistical metrics like R-squared.

Main Results:

• System identification effectively models the emotional trajectory of music.
• Achieved an average R2 of 21.9% for valence and 78.4% for arousal.
• Demonstrated generalization of emotional content across a music genre.

Conclusions:

• System identification offers a powerful approach to modeling music's emotional dynamics.
• The proposed methodology enhances previous time-series models and reduces classification ambiguity.
• This research provides a generalized framework for understanding music-induced emotions.