Just noticeable difference for simulation accuracy between full and reduced order models (L)

Abstract

Model order reduction techniques significantly reduce the computational time when performing accurate room acoustic simulations with numerical methods that inherently include all the wave phenomena. There is a clear trade-off between physical accuracy and acceleration, but how humans perceive these errors is unknown. This study aims to investigate physical error limit that does not induce perceptual differences. Various two-dimensional rooms and reverberation times are tested with a three-alternative forced-choice listening test. Results reveal that for the presented cases, the threshold stands between a relative root mean square error of 1% and 0.1%, where the reduced order model stimulus results in a statistically significant difference.