Familiar Feelings: Listener-Rated Familiarity in Music Emotion Recognition

Abstract

The task of music emotion recognition (MER) has previously been explored using a variety of audio, lyrical, and basic meta-data features. As feature extraction and classification algorithms advance, the need for relevant extra-musical features becomes apparent. The efficacy of familiarity as a feature in a MER system was evaluated by a Random Forest feature importance analysis on a novel dataset of 5000 clips with annotated familiarity and valence. Familiarity was correlated to perceived valence (r = 0.250) and resulted in a statistically significant increase of 0.011 in the F-score of a baseline MER classifier upon its inclusion.