Emotion-based Music Analysis and Beyond

Automatic recognition of the perceived emotion of music allows users to retrieve and organize their music collections in a fashion that is content-centric and intuitive. A typical approach to music emotion recognition categorizes emotions into a number of classes and applies machine learning techniques to train a classifier. This approach, however, faces a granularity issue that the number of emotion classes is too small in comparison with the richness of emotion perceived by humans. In this talk, I would introduce some research that takes a very different perspective and views emotions as points in a 2-D space spanned by two latent dimensions: valence (how positive or negative) and arousal (how exciting or calming). In this approach, music emotion recognition becomes the prediction of the valence and arousal values of a song corresponding to a point in the emotion plane. This way, the granularity and ambiguity issues associated with emotion classes no longer exist since no categorical class is needed. Moreover, because the 2D plane provides a simple means for user interface, new emotion-based music organization, browsing, and retrieval can be easily created for mobile devices that have small display area.