An exploratory study of population differences based on massive database of physiological responses to music

Understanding the relationship between music and physiological indicators of emotion is a critical part of affective computing. However, the limitations of small sample sizes make it hard to draw convincing and practical conclusions. In this paper, we extend an unsupervised clustering method to a massive music-electrophysiological dataset which contains 6,509 valid samples. After clustering the people who have similar electrodermal activity (EDA) responses to music, we apply Chi-square test and ANOVA to the questionnaires associated with the EDA responses. The results reveal the impact of demographic factors (i.e. gender, age, and nationality) on EDA responses, as well as the relationships between self-reported emotional responses and bodily EDA responses to music at a much higher confidence level than previous studies.