Temporal Integration of Emotion Perception for Cross-Cultural and Multi-Emotion Music

Music can bring different emotional feelings which often change with time. The whole emotion-perception of a piece of music is related to the emotion of each temporal segment obviously. This study focuses on the relationship between the whole and each segment in music emotion-perception and how does the relationship change under different culture as well as different mood categories. Specifically, we built a temporal-dynamic emotion dataset containing Chinese folk music and Western classical music, and the music was also divided into four categories: sad, angry, happy and relaxed. Every piece of music is divided into several temporal segments of different lengths. Valence-Arousal model was used to annotate the emotion of segments and the whole music. Through correlation analysis and regression analysis, the emotion-perception relationship between the whole and segments was studied. The results show that relationship between the whole and each segment in emotion-perception is significant. This paper also explored the differences of this relationship under different cultural backgrounds and emotional categories.