Finding an ideal level of syncopation to elicit a groove response

Groove has often been described as a sensorimotor coupling with music, where the listener feels the urge to engage and move along with the music. Much recent work to measure the level of groove in music relies on listener ratings. These studies suggest that metrical syncopation, micro-timing, and spectral flux may be important features in high-groove music. In this paper we use an established list of 80 musical works, ranked by listener ratings on their level of groove, as a basis for correlating the presence of syncopation to their perceived level of groove. Specifically, we use Madmom, a machine learning method, to extract metrical structure from the recordings of these musical works. Next, we apply to these metrical structures a novel approach to Inner Metric Analysis (IMA) as a means for measuring their levels of syncopation. IMA generates a metric weight profile of a piece based on the weight assigned to each note onset, calculated by computing each onset’s relation to the pulse stream, and returns a measure of syncopation based on its entropy value. We then correlate the metric weight profiles with the listener groove ratings using a regression model. We expect to find that both low and high levels of syncopation will correspond to low groove ratings, while moderate levels of syncopation correspond to higher groove ratings, creating an inverted U-shaped curve. These results will allow us to suggest that ideal levels of syncopation fit within a quantifiable entropy distribution. Correspondingly, this model will also allow us in the future to make more comparisons between theoretic and perceptual studies, establishing new ways to understand how listeners interact with music.