Probing Amplitude-Modulation Detection and Phoneme Categorization with Auditory Reverse Correlation - Archive ouverte HAL Accéder directement au contenu
Poster De Conférence Année : 2022

Probing Amplitude-Modulation Detection and Phoneme Categorization with Auditory Reverse Correlation

Résumé

It is now acknowledged that part of the deleterious effect of a steady noise on the auditory perception of amplitude-modulation (AM) or speech sounds is due to the masking of useful modulations in the signal by the intrinsic random envelope fluctuations arising from the filtering of noise into cochlear critical bands (Dau et al., 1997, 1999). Reverse correlation techniques are particularly suitable for exploring the confounding effect of noise on perception (Murray, 2011), in particular for auditory perception (Varnet et al., 2013). However, the only attempt at applying this technique to an AM-detection-in-white-noise task yielded mixed results (Ardoint et al., 2007). Here, we analyzed the effects of a white noise masker on AM detection, in order to unveil the subject’s decision strategy in the task. Participants were required to detect the presence of a 4-Hz sinusoidal AM target applied to a 1-kHz tone carrier embedded into a white noise. Then, using a reversed-correlation analysis on the collected data, we were able to reveal which aspects of the temporal envelope of the noises influenced the listener’s responses. These results where compared with simulated data from standard modulation-perception models. In a second study, we employed a similar approach on a logatome categorization task in noise ("aba" vs. "ada"), revealing the acoustic cues critical for the identification of phonemes /b/ and /d/. These two studies offer a direct illustration of the role of intrinsic modulations of noise maskers on auditory perception in noise: At a given signal-to-noise ratio, some noise samples have a greater influence than others depending on the exact temporal distribution of envelope energy. We also demonstrate that this effect of noise can be used to probe auditory perception in noise in real and simulated listeners.
Fichier non déposé

Dates et versions

hal-03848386 , version 1 (10-11-2022)

Identifiants

  • HAL Id : hal-03848386 , version 1

Citer

Léo Varnet, Christian Lorenzi, Alejandro Osses. Probing Amplitude-Modulation Detection and Phoneme Categorization with Auditory Reverse Correlation. 16ème Congrès Français d'Acoustique, CFA2022, Apr 2022, Marseille, France. ⟨hal-03848386⟩
36 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More