Exploring Roughness Perception in Car Engine Noises through Complex Cepstrum Analysis

Abstract

Car engine noises have been analyzed from a psycho-acoustical point of view since early 90's. Classical psycho-acoustic parameters as loudness, sharpness and roughness have been used for this task, resulting in significant improvements in the annoyance caused in humans, especially for diesel engine noises. However, roughness sensation has been well defined and measured in modulated narrowband signals, but not in broadband noises. Here a new model of engine noise production is proposed in order to relate some of its features with the roughness perceived. The model is based on complex cepstrum analysis of the real engine noise, a technique widely used in speech processing, but not explored yet for psycho-acoustic sensations. A noise synthesizer is build for performing subjective tests and detect model parameters related to roughness perception. Results have shown that peak amplitudes range (represented by a certain amplitude modulation index of the impulses) is linearly related to roughness perceived, whereas frequency modulation index of the same impulse train has a weak influence on roughness sensation.