MODELING MUSICAL MOOD FROM AUDIO FEATURES AND LISTENING CONTEXT ON AN IN-SITU DATA SET

Real-life listening experiences contain a wide range of music types and genres. We create the first model of mu-sical mood using a data set gathered in-situ during a us-er’s daily life. We show that while audio features, song lyrics and socially created tags can be used to successful-ly model musical mood with classification accuracies greater than chance, adding contextual information such as the listener’s affective state or listening context can improve classification accuracy. We successfully classify musical arousal with a classification accuracy of 67 % and musical valence with an accuracy of 75 % when using both musical features and listening context. 1