Impression-Based Music Playlist Generation Method for Placing the Listener in a Positive Mood

Abstract

This paper proposes an impression-based music playlist generation method with musical diversity and serendipity to positively impact the mood of a listener. The method designs two types of playlists with smooth transitions for audio track impressions: mood-boosting (uplifting) and mood-stabilizing (relaxing). Both these impressions have four patterns with different structures that alter the listener’s mood. To create a playlist, the proposed method calculates the probability of impressions for all audio tracks using a multinomial mixture model adopting a Bayesian approach and chooses audio tracks based on these probabilities. This study uses two psychometric evaluations to evaluate sample playlists for all four patterns of both playlists, i.e., whether they can positively influence a Japanese listener’s mood. The results indicate that the proposed method can achieve both these objectives concerning the audio tracks.