by Kees van den Doel
We describe and test methods to construct modal resonance models for solid objects, suitable for the real-time synthesis of sound-effects in simulation and animation. Measurements on typical everyday objects such as a metal vase or a bowl result in several hundred modes, of which only a small fraction is perceptually relevant. We have proposed several heuristics, inspired by psycho acoustical data, to select the modes by perceptual relevance and to order them so that one can increase the quality by adding more modes, at the price of additional computational complexity (progressive synthesis). The resulting synthetic sounds are tested on human subjects in order to determine the quality of the sounds relative to the target sound which they are designed to approximate. The resulting data is used to verify and tune the mode selection methodologies, and to increase our understanding of what determines the subjective quality of a synthetic sound effect.