Numerical comparison of different equivalent source models and source quantification techniques for use in sound synthesis systems

摘要

The aim of sound synthesis methods is to auralize the sound of a physical sound source in the way it is perceived by a listener at a particular location. Most existing synthesis methods start with a source-transmission path - receiver (s-t-r) model to predict spectral characteristics of the sound at the receiver location (e.g. a time-frequency spectrum). To define such a s-t-r model, an adequate representation of the physical sound source is crucial for achieving high quality sound syntheses. In the mid-frequency range, equivalent source models composed of a limited number of elementary substitute sources are usually applied. These equivalent models overcome many of the disadvantages of other sound source reconstruction models. However, quantification of the elementary substitute sources is difficult. After propagation along the transmission path and based upon the predicted spectral characteristics, the sound synthesis methods then synthesize an audible sound at the receiver location. This paper presents a numerical study on the performance of different equivalent source models and quantification techniques. It is shown that the recently developed inverse quantification approach based on acoustic particle velocities can be a fast and sufficient accurate alternative for techniques based on velocity and acceleration on the source surface or on acoustic pressures at indicator positions around the source. Furthermore, guidelines are given to select the number of equivalent substitute sources.