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

Naim Mansour,Tobias May,Marton Marschall,Adam Westermann,Torsten Dau

doi:10.1121/10.0008980

Abstract

Measures of "aided" speech intelligibility (SI) for listeners wearing hearing aids (HAs) are commonly obtained using rather artificial acoustic stimuli and spatial configurations compared to those encountered in everyday complex listening scenarios. In the present study, the effect of hearing aid dynamic range compression (DRC) on SI was investigated in simulated real-world acoustic conditions. A spatialized version of the Danish Hearing In Noise Test was employed inside a loudspeaker-based virtual sound environment to present spatialized target speech in background noise consisting of either spatial recordings of two real-world sound scenarios or quadraphonic, artificial speech-shaped noise (SSN). Unaided performance was compared with results obtained with a basic HA simulator employing fast-acting DRC. Speech reception thresholds (SRTs) with and without DRC were found to be significantly higher in the conditions with real-world background noise than in the condition with artificial SSN. Improvements in SRTs caused by the HA were only significant in conditions with real-world background noise and were related to differences in the output signal-to-noise ratio of the HA signal processing between the real-world versus artificial conditions. The results may be valuable for the design, development, and evaluation of HA signal processing strategies in realistic, but controlled, acoustic settings.

Highlights

Hearing aids (HAs) attempt to restore hearing-impaired (HI) people’s ability to reliably explore their auditory world
The results of the present study suggest that a realistic virtual sound environments (VSEs) can provide an effective setting for evaluating the impact of hearing aids (HAs) signal processing algorithms on a listener’s speech intelligibility (SI) performance and for relating that performance to instrumental HA performance metrics
The effect of HA dynamic range compression (DRC) processing on SI was investigated in two realistic acoustic scenes, constructed using spatial background recordings and anechoic speech samples convolved with room impulse response (RIR) measured in situ

Summary

Introduction

Hearing aids (HAs) attempt to restore hearing-impaired (HI) people’s ability to reliably explore their auditory world. HAs have failed to provide a consistent SI benefit across users (Kochkin, 2002). This may be partially due to the focus of current HA fitting procedures on restoring audibility, rather than addressing supra-threshold distortions which HI listeners commonly experience when listening to speech in noisy situations [e.g., Sanchez-Lopez et al (2019)]. Signal processing algorithms in HAs have mostly been optimized for SI using speech-recognition-in-noise metrics, such as speech reception thresholds (SRTs), obtained with artificial acoustic stimuli, which may not correlate well with HA satisfaction in the listeners’ real-world experience (Bentler et al, 1993; Cord et al, 2007; Wu, 2010). It may be worthwhile to explore SI in more realistic, ecologically valid ways, both in unaided conditions as well as in conditions aided by the HA

Objectives

Methods

Findings

Discussion

Conclusion