Contribution
Audio-visual scene analysis in aided hearing-impaired listeners using virtual reality
* Presenting author
Abstract:
Despite improvements in signal-to-noise ratio (SNR) shown by state-of-the-art hearing aids (HAs) in conventional measurements, users still report challenges in real-life situations. This suggests that such measurements may not fully capture real-world listening environments. This study proposes a more ecologically valid paradigm for perceptually evaluating HA performance using virtual reality technology. A spherical loudspeaker array was used for sound reproduction, paired with a head-mounted display for visual presentation. The paradigm includes two tasks: speech localization and speech comprehension. During the localization task, aided hearing-impaired participants could move their heads to localize the target talker positioned within the frontal horizontal plane, among interfering talkers and noise. During this task, response time and localization errors were measured. In the comprehension task, participants listened to a newsclip, followed by a related binary question. Results showed that response time, localization errors and error rates in answering binary questions increased as the number of interfering talkers rose from 1 to 3. Localization accuracy improved with higher noise reduction settings in a 4-talker scenario, while no significant trends were observed in a 2-talker scenario. This new paradigm enables HA performance assessment in situations that are more closely aligned with everyday listening experiences.