Speech intelligibility in a realistic virtual sound environment

Abstract

Employing realistic, yet controlled sound scenarios for the evaluation of hearing aid (HA) algorithms in a virtual sound environment (VSE) has the potential to improve HA users' real-world listening experience and performance. In the present study, so-called critical sound scenarios (CSS) were defined as acoustic scenes that HA users experience as important, difficult and common, and selected through ecological momentary assessment (EMA), inspired by results from recent studies. The scenarios were acquired in real scenes using a spherical microphone array, and reproduced in an acoustically valid way inside the VSE. A speech intelligibility task using the Danish Hearing In Noise Test (HINT) was implemented in one of the scenes to measure speech reception thresholds (SRT) for normal-hearing and hearing-impaired subjects. SRTs obtained for the realistic VSE background noise were found to be higher than corresponding ones reported in headphone-based studies, potentially as a result of the increased cognitive effort required to separate target speech from the more realistic interfering speech background.

The effect of hearing aid dynamic range compression on speech intelligibility in a realistic virtual sound environment

“…It was found that both unaided and aided SRTs were significantly increased inside both realistic VSEs compared to a reference condition employing quadraphonic SSN as a masker. This is consistent with previous studies, which showed increased spectral and modulation-spectral energetic masking for realistic noises as well as a detrimental effect of intelligible speech in the maskers on SI (Best et al, 2015;Mansour et al, 2019). Despite the acoustic differences, SI was similar between the two realistic noise types, suggesting that precise acoustic details in the reproduced environment may be less important than its overall loudness and number of interfering sources.…”

“…The main aim of this study was to investigate the impact of applying in situ background-noise recordings as maskers in a speech intelligibility task conducted with and without hearing aids. Consistently with previous work (Best et al, 2015;Mansour et al, 2019), employing realistic noise maskers resulted in increased SRTs compared to artificial noise maskers. This reduction in SI in realistic backgrounds containing a mixture of interfering talkers has been linked to energetic speech-on-speech masking of the interferers on the target speech in the spectral and modulation-spectral domains (Brungart et al, 2006;Jørgensen and Dau, 2011) as well as informational masking caused by the intelligibility of the interferers (Westermann and Buchholz, 2015).…”

“…Best et al (2015) found that HAs provided a greater benefit to SI in their simulated VSEs compared to more artificial masker types for the majority of listeners. Here, the effect of HA processing was evaluated inside recorded VSE using Ambisonic auralizations of spatial recordings (Mansour et al, 2019). Two real-world scenes were recorded with a spherical microphone array using higher-order Ambisonics and reproduced inside a 64-loudspeaker loudspeaker array, providing the background-noise masker for the VSE.…”

Measuring Speech Intelligibility and Hearing-Aid Benefit Using Everyday Conversational Sentences in Real-World Environments

“…This means that by selecting the right environment a useful operating point (where scores are away from both ceiling and floor) can be found for listeners across a wide range of hearing abilities. Depending on the specific purpose, the test environment may be selected based on the individual’s hearing loss, their reported speech-in-noise problem, or the relevance of a test environment (e.g., see Mansour et al, 2021 ). Furthermore, we demonstrated that, within our framework, the choice of speech materials not only affected the realism of the stimuli but also changed the difficulty of the listening task.…”

Attention to Speech and Music in Young Children with Bilateral Cochlear Implants: A Pupillometry Study

“…Additionally, we also recreated a virtual 3D acoustic environment to make the task more ecologically valid. Such virtual acoustic realities have been recently shown to increase the difficulty of the listening task compared to lab-based assessments due to the added complexity, especially in patients with HI [ 65 , 66 ]. In the presence of a virtual 3D acoustic environment, it is possible that the intelligibility decreased excessively in both high and low background noise, and that young children with CIs subsequently found the task of listening to speech too demanding when background noise was present.…”