MODIFIED AURES'S MODEL FOR ACOUSTIC ROUGHNESS COMPUTATION BASED ON A BANK OF DIGITAL FIR FILTERS

Stanislav Marchevský - Dušan Kocur - Ladislav Mihalčík - Pavol Galajda - Miloš Drutarovský

   The paper deals with a design of MATLAB software tools for acoustic roughness computation of digitized sound signals based on modified Aures's models. The original Aures's model [1] assumes the digitized sound signals sampled with sampling frequency f_s=40960 Hz. In this paper, two modifications of Aures's model for the roughness computation will be proposed. The first modification allows to extend the Aures's model for the roughness computation also for a direct computation of sound signal roughness if sampling frequency is different from f_s=40960 Hz. The second modification is represented by the substitution of the experimentally determined weighted functions applied for the determining of relevant envelope fluctuations of analyzed sound signal by appropriately designed digital filters. This modification enables also to filter the envelope fluctuations in time domain by means of an aperiodic discrete convolution. It enables to decrease the computational complexity of Aures's model for roughness computation. The proposed methods were tested on sound signals prepared by Technical Acoustics Department of Technical Testing Institute in Piešťany. The test sound signals were sampled with sampling frequency f_s=25 kHz. The obtained results confirm that the proposed modifications of the roughness computation models can provide results similar to that of original Aures's model.

Keywords: psychoacoustics, roughness, FIR filters, Aures's model, discrete convolution, MATLAB

[full-paper]