Deuxièmes Journées Perception Sonore 2012

10 et 11 décembre 2012

CNRS Laboratoire de Mécanique et d'Acoustique

31, Chemin Joseph Aiguier

13009 Marseille


Temporal Entropy-Based Texturedness Indicator for Audio Signals: Short Time Perceptual Analysis.


Olfa FRAJ, Raja GHOZI et Mériem JAIDANE

Unité de recherche Signaux et Systèmes U2S, Dép. Technologies de l’Information et des Communications, Ecole Nationale d’Ingénieurs de Tunis (ENIT), BP37 - Le Belvédère - 1002 Tunis,

(This paper has been submitted to IEEE Transactions on Audio, Speech and Language Processing, October 2012)
In this paper, we present a temporal entropy-based indicator that reflects the texturedness level of a given audio signal. The proposed indicator is based on a direct and reverse signal entropy analysis, through a progressive auditory information content tracking. A [0 - 5] texturedness indicator is then constructed using perceptually-inspired auditory parameters, and is inherently associated with short listening time. The robustness of the proposed indicator to small variations of its computational parameters is tested. Using this indicator, a new audio signals classification is proposed where speech signals are assigned low texturedness scores, and academic noise signals are assigned the higher range of texturedness scores. Classically known audio textures are assigned, in this scale, various intermediate to high texturedness scores.
Index Terms - Audio texturedness indicator, audio textures, audio signals classification, direct and reverse cumulative entropy, perceived auditory information, short listening time.
Note: This work has been supported by the Tunisian French CMCU project, titled: “ Altered Perceptions in Urban soundscapes: An Audio Texture Approach: Analysis and Corrections”, among U2S-ENIT, ERA- ENAU, CERMA and CRESSON.
U2S-ENIT: Research Unit on Signals and Systems, National School of Engineering, Tunis, Tunisia
ERA-ENAU: Research Group on Ambiances of the National School of Architecture and Urbanism, Tunis Tunisia