Synthesis of Perceptually Plausible Multichannel Noise Signals Controlled by Real World Statistical Noise Properties

摘要

In this paper we propose an algorithm to generate artificial but perceptually plausible acoustic noise disturbances. The method enables to define desired noise properties like coloration, modulations, and amplitude distribution independently of each other while also providing means to define a desired coherence between signal channels. These properties can, for instance, be extracted from a short exemplary noise recording to then synthesize a noise signal of arbitrary duration with the same characteristics. Results of a listening test in a monaural setup show no significantly different naturalness between synthesized and original signal for noise types of technical origin.