Modeling formant-frequency discrimination based on auditory-nerve and midbrain responses: normal hearing and sensorineural hearing loss

U-Cheng Leong1, Douglas M. Schwarz2, Joyce M. McDonough3, Laurel H. Carney2,4

1 Departments of Otolaryngology, University of Rochester, Rochester NY, USA; 2 Departments of Neuroscience, University of Rochester, Rochester NY, USA; 3 Departments of Linguistics, University of Rochester, Rochester NY, USA; 4 Departments of Biomedical Engineering, University of Rochester, Rochester NY, USA

Background: Formant-frequency difference limens (FFDLs) are increased in listeners with sensorineural hearing loss (SNHL) when compared to normal hearing (NH) controls. Preliminary results show that thresholds of the SNHL group is more affected by broadening the formant bandwidths (BW). We tested the hypothesis that response profiles of neurons in the inferior colliculus (IC), based on the neural fluctuation model (Carney, 2018), could predict FFDLs in listeners with NH and SNHL. 

Methods: Behavioral thresholds were estimated in a two-down-one-up, four-interval, two-alternative-forced-choice paradigm. Computational models for auditory-nerve (AN, Zilany et al., 2014) and IC neurons (Mao et al., 2013) were used to simulate response profiles across neural populations (characteristic frequency=150-3000 Hz). FFDLs were estimated based on the Mahalanobis distances between model response profiles to each stimulus interval and a template based on the standard stimulus. Noise in the simulations was based on spontaneous activity of AN fibers. Model thresholds were compared to the behavioral results.

Results: FFDLs were estimated for seven listeners by including pure-tone thresholds in the peripheral model. Trends across BW and different degrees of SNHL agree with behavioral results. Threshold estimates based on AN population responses were generally higher than behavioral thresholds, while IC-based thresholds were lower. Thus IC-rate profiles can potentially explain human thresholds for this task.

Conclusion: The formant-frequency discrimination task is an alternative to intelligibility testing for quantifying sensitivity to changes in speech sounds in listeners with SNHL. Thresholds for all listeners increased with BW, as predicted by computational models of AN and IC rate profiles. IC responses are driven by neural fluctuations; therefore, results suggest strategies for enhancing intelligibility by manipulating neural fluctuations. (NIH-R01-DC001641)

Cheng Leong et al
Figure 1. Stimulus spectra of standard (blue) and target (orange) vowel-like stimuli with ΔF2=2% and the corresponding model responses of AN fibers and IC band-enhanced cells (BE), which are excited by fluctuations within a band of modulation frequencies.