Virtual sound generation by linear modal synthesis based on recorded audio examples

Abstract

Linear modal synthesis (LMS) has been investigated to synthesize virtual sound of rigid bodies. The existing methods to synthesize realistic sound of virtual objects lack the real-time computations of perceptually satisfactory material parameters. In this study, a novel approach is used for the estimation of the material parameters that record the intrinsic quality of interacting materials and to extract significant features from the recorded impulse responses. A parameter estimation technique based on psychoacoustic principles is developed that incorporates material features to search the best material parameters for LMS. The synthesized audio produced from LMS shows the same perception of material as a recorded sound.