String Instrument Body Modeling Using FIR Filter Design and Autoregressive Parameter Estimation

Friedrich v. Türckheim; Thorsten Smit; Robert Mores
DAFx-2010 - Graz
In this paper, a method is presented for modeling string instrument bodies. The crucial point of the method is a two-step filtering technique which combines advantages of both finite impulse response (FIR) filter design and autoregressive (AR) modeling. The frequency sampling method for FIR filter design enables modeling of specific resonance constellations within the magnitude spectrum, whereas a subsequent all-pole modeling step reduces the filter order. In addition, frequency warping is applied in order to further decrease the model complexity. The proposed method allows for highly-detailed modifications of individual resonances without affecting other resonances. An example for modifying a virtual violin body is presented as well as the implementation on a real-time platform which allows for experiments on perceived violin sound and musician-instrument interaction.