Analysis/synthesis of bowing control applied to violin sound rendering via physical models

Abstract

A prominent challenge in instrumental sound synthesis is to reproduce the expressive nuances naturally conveyed by a musician when controlling a musical instrument. Despite the flexibility offered by physical modeling synthesis, appropriately mapping score annotations to sound synthesis controls still remains an interesting research problem, especially for the case of excitation-continuous instruments. Here we present our work on modeling bowing control in violin performance, and its application to sound synthesis via physical models. Minimally invasive sensing techniques allow for accurate acquisition of relevant timber-related bowing control parameter signals. The temporal contours of bowing control parameters (bow velocity, bow force, and bow–bridge distance) are represented as sequences of low-order polynomial curves. A database of parametric representations of real performance data is used to construct a generative model able to synthesize bowing controls from an annotated score. Synthetic bowing controls are then used to render realistic performances by driving a violin physical model.