Video Visualization of Predictors of Emotions Dynamically Expressed by Music

摘要

Music is particularly appreciated for its capacity of evoking a large range of emotions. A growing area of research in music psychology and computer music endeavors to understand and model the role of the different constituents of music in its overall emotional impact. Whereas research so far has mainly focused on global musical and emotional descriptions, our project investigates more in detail the relationship between the dynamic evolution of musical content and the dynamic development of the resulting emotional reaction. We introduce a new method for predicting the dynamic appreciation of emotion based on audio and musical descriptions. We focus on one particular emotion: power. We propose a predictive model of the dynamic evolution of power based on several components: energy of sustained sound, pitch activity and tonal steadiness. We also introduce a new framework for visualizing, for any piece of music, the relationship between the dynamic emotion ratings and the audio and musical content playing a role in the perceived emotion.