The Role of Time in Music Emotion Recognition

Music is widely perceived as expressive of emotion. Music plays such a fundamental role in society economically, culturally and in people’s personal lives that the emotional impact of music on people is extremely relevant. Research on automatic recognition of emotion in music usually approaches the problem from a classification perspective, comparing “emotional labels” calculated from different representations of music with those of human annotators. Most music emotion recognition systems are just adapted genre classifiers, so the performance of music emotion recognition using this limited approach has held steady for the last few years because of several shortcomings. In this article, we discuss the importance of time, usually neglected in automatic recognition of emotion in music, and present ideas to exploit temporal information from the music and the listener’s emotional ratings. We argue that only by incorporating time can we advance the present stagnant approach to music emotion recognition.