Audio Features for Music Emotion Recognition: A Survey
This article presents a survey on the existing emotionally-relevant computational audio features, supported by the music psychology literature on the relations between eight musical dimensions and specific emotions.
Abstract
The design of meaningful audio features is a key need to advance the state-of-the-art in music emotion recognition \n(MER). This article presents a survey on the existing emotionally-relevant computational audio features, supported by the music \npsychology literature on the relations between eight musical dimensions (melody, harmony, rhythm, dynamics, tone color, \nexpressivity, texture and form) and specific emotions. Based on this review, current gaps and needs are identified and strategies for \nfuture research on feature engineering for MER are proposed, namely ideas for computational audio features that capture elements \nof musical form, texture and expressivity that should be further researched. Previous MER surveys offered broad reviews, covering \ntopics such as emotion paradigms, approaches for the collection of ground-truth data, types of MER problems and overviewing \ndifferent MER systems. On the contrary, our approach is to offer a deep and specific review on one key MER problem: the design of \nemotionally-relevant audio features.