Connecting a Biophysical Auditory Periphery Model to Perceptual Back-ends for Psychoacoustic Performance Prediction across Tasks - Archive ouverte HAL Accéder directement au contenu
Poster De Conférence Année : 2020

Connecting a Biophysical Auditory Periphery Model to Perceptual Back-ends for Psychoacoustic Performance Prediction across Tasks

Alejandro Osses
Sarah Verhulst
  • Fonction : Auteur
  • PersonId : 1185392

Résumé

Biophysical models of the human auditory periphery can simulate objective markers of peripheral hearing such as otoacoustic emissions, auditory brainstem responses (ABRs), and envelope-following responses (EFRs), and can be used to simulate how sensorineural hearing loss affects these responses. Given their physiological detail, it would be interesting to extend these models to include a perceptual back-end and allow a prediction of behavioral detection thresholds. In the most basic approach, simulated neural responses to reference and target stimuli in a three-alternative forced-choice (3-AFC) psychoacoustic tasks can be compared to each other to yield a detection cue. The detection cue degradation as a function of outer-hair-cell deficits or synaptopathy can inform how these hearing deficits impact detection thresholds. This basic approach also predicts less accurate responses and degraded thresholds when the external stimulus variability increases. In this work, we compare different psychoacoustic back-ends to evaluate which approach more accurately captures behavioral thresholds of normal and hearing-impaired listeners across a range of detection tasks in quiet or background noise. We simulated an amplitude-modulation detection task with tonal and noise carriers (fc=4000 Hz) which were either presented in quiet, or in a broadband or narrowband noise masker. The considered psychoacoustic back-ends included a within-trial RMS detector, a classical template-based optimal detector, and a template-based RMS detector. In our analysis, we compared simulated and experimental thresholds across N=21 listeners with normal or high-frequency sloping audiograms. Simulated detection cues were analyzed across the different cochlear characteristic frequencies and we discuss how simulated detection thresholds are impacted by (1) different sensorineural hearing loss profiles, and (2) different inherent properties of the background noises (external variability). Although within-trial model back-ends are more directly related to the task, we observed a higher robustness of template-based approaches to external stimulus variability. This supports the view that listeners are able to “internalize” both target and reference sounds over the course of a task and speaks for the inclusion of basic top-down approaches in auditory perception models.
Fichier principal
Vignette du fichier
2020_Osses2020d-Verhulst-ARO.pdf (1.43 Mo) Télécharger le fichier
Origine : Fichiers éditeurs autorisés sur une archive ouverte

Dates et versions

hal-03873557 , version 1 (24-02-2023)

Identifiants

  • HAL Id : hal-03873557 , version 1

Citer

Alejandro Osses, Sarah Verhulst. Connecting a Biophysical Auditory Periphery Model to Perceptual Back-ends for Psychoacoustic Performance Prediction across Tasks. ARO Midwinter Meeting, Jan 2020, San José, United States. ⟨hal-03873557⟩

Collections

TDS-MACS
24 Consultations
3 Téléchargements

Partager

Gmail Facebook X LinkedIn More