The Speech Envelope Following Response in Normal and Hearing Impaired Listeners

Abstract

The aim of this work was to study the perceptual relevance of the frequency following re-sponse to the syllable /da/ for speech intelligibility in noise based on age and hearing defi-cits. Recordings of the auditory evoked potential from 44 young and older individuals were analyzed. The envelope following responses were analyzed in terms of amplitude, latency and robustness to noise. The subcortical response was also simulated to form predictions on the effect of cochlear synaptopathy and outer hair cell loss on the EFR. The results corroborate the observation that the EFR is reduced as a consequence of ageing and high-frequency hearing loss. Both the audiogram and size of the speech EFR fell short in the prediction of speech reception thresholds in noise, but they accounted well for group performances. We introduce a method to obtain EFR latency with a cross covariance matrix. Validation of the method confirmed that the variance in speech EFR latency matched the variance in click ABR Wave V peak latency, but the match was better for the younger subjects. The influence of noise on the EFR was also assessed. The robustness to noise depended on HF hearing in the older normal hearing population, but was indicative of EFR amplitude in the younger adults.

Differences between children and adults in the neural encoding of voice fundamental frequency in the presence of noise and reverberation

“…A second piece of evidence displaying the simplified nature of the central processing is the reduced amplitude of the simulated EFRs across all conditions (Figure 6a). This was evident in the current study despite previous reports of the model producing EFRs of similar amplitude to those experimentally recorded in normal hearing individuals when presented with either amplitude-modulated stimuli (Vasilkov et al, 2021) or a non-modified vowel (/da/, f 0 $120 Hz, Wartenberg et al, 2022). The most obvious explanation for our modelling results is that the weighting factors used to normalize brainstem/midbrain neural activity do not take into account our frequency-specific approach restricting energy associated with the F1 and F2+ formants to the low (F1) and high (F2+) frequency channels in the model: the resulting linear summation of neural activity from these parallel pathways in the AN, CN, and IC stages ultimately leading to smaller F1 and F2+ components in the simulated EFRs.…”

Model-based hearing-enhancement strategies for cochlear synaptopathy pathologies

“…The EFR magnitudes between the two groups also showed statistically significant differences ( p = 0.0005 for the unprocessed condition). Since the speech-evoked EFR reflects the neural encoding of the pitch information of speech [72, 80], the decreased EFR magnitudes in the older group suggest that pitch tracking was degraded by age-related CS. Although the oNH EFR magnitudes were improved after processing, the (10,0,0) processed speech stimuli were not able to fully restore the oNH magnitudes to the yNH level.…”

“…To generalise the two processing types for different audio stimuli, we first focussed on high-pass (HP) filtered speech (i.e., speech content above f cut = 1.65 kHz) which relates more to the SAM stimuli. The intelligibility of HP-filtered speech relies on the coding of the temporal envelope of sound [68,69] and has been linked to the EFR results of listeners in previous CS-related studies [62,[70][71][72]. After establishing the envelope processing for HP-filtered speech and evaluating the simulated restoration that was achieved, the CS-compensating algorithms can be used in exactly the same way to process any stimulus.…”