Music emotion recognition (MER) as a part of music information retrieval (MIR), examines the question which parts of music evoke what emotions and how can they be automatically classified. Classification systems need to be trained in terms of feature selection and prediction. Due to the subjectivity of emotions, the generation of appropriate ground truth data poses challenges for MER. This paper describes obstacles of defining and measuring emotions evoked by music. Two methods, in principle able to overcome problems in measuring affective states induced by music, are outlined and their results are compared. Although the results of both methods are in line with psychological theories of emotions, the question remains how good the perceived emotions are captured by either method and if these methods are sufficient for ground truth generation.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.