Computer-based music feature analysis mirrors human perception and can be used to measure individual music preference

This paper explores the measurement of individual music feature preference using human- and computer-rated music excerpts. In the first of two studies, we correlated human ratings of song excerpts with computer-extracted music features and found good accordance, as well as similar criterion validity with preference for musical styles (the MUSIC model, mean r = 0.88). In a second online study (N = 2118), using PCA and Procrustes analysis, we found that measured music preference showed the same established three-component structure from previous research (Arousal, Valence, Depth), regardless of whether the music pieces were rated by humans or the ESSENTIA music analysis software. Our results suggest that computer-extracted music features can be used to assess individual music preference.