Optimization of sound synthesis model parameters

In this document we present a sound synthesis model prototype where a realistic sound reproduction of acoustic musical instruments is the final goal. A physically informed digital waveguide synthesis method was chosen for modelling the instrument properties. The method was applied for the case of the piano. A procedure was presented for design, analytical derivation and automatic calibration of the synthesis model parameters. A procedure for calibration with multiobjective minimization of perceptual differences between the sources and synthesized sounds was proposed. Four perceptually based sound feature distances with fast computation capabilities were chosen for optimization. For parameter search, a differential evolution scheme for multiobjective optimization was used. Finally, extension possibilities of the calibration method and future directives are shown