A METHOD FOR COMPUTER CHARACTERIZATION OF "GESTURE" IN MUSICAL IMPROVISATION
In the design of interactive computer music systems and the composition of interactive computer music, the tracking and analysis of musical "gestures" — characteristic motions discerned within musical attributes — provides a promising challenge. There are in fact ways that one can clearly and empirically define and identify "gesture" in musical content, often with conceptual models and tools similar to those used fortracking and identifying physical gestures. The analysis of musical gesture as "significant motion" can be applied to many aspects of music: melodic contour, notespeed and density, loudness, level of dissonance, etc. Gestures can be characterized by the shapes producedby measuring changes in these aspects, and the derivation of data about change, rate of change, etc. within a particular feature or set of features.
Computer evaluation of gesture may be divided into the tasks of measurement, segmentation, identification, andtaxonomy. What are the elements of musical gesture and how can a computer best discern them? How can a computer know when a gesture begins and ends? Howcan different, unforeseen gestures be compared and classified? Perhaps most significantly, how can a computer, once it has identified and characterized a gesture, attribute musical meaning to it? This research proposes criteria and groundwork for the tracking, measurement, and analysis of "gesture" in the musical content of sound structure, and the use of that analysisin interactive computer music.