Predicting Perceived Emotions in Music: the Impact of Genre

In this work, we assessed the impact of using genre information in the automatic classification of perceived emotion in music. In this process, we developed a dataset in which tracks were mapped according to their genre and their perceived emotion. Our results show that using a specific classifier for each genre yields better results than using a single classifier, with no genre information, for Indie-Rock, Jazz, Heavy-Metal, and Classical music. However, classification result were poor for Bossa-Nova. Therefore we speculate that Bossa-Nova conveys emotion differently than the other tested genres.