P53Session 1 (Thursday 11 January 2024, 15:35-18:00)Influence of semantic context information on rollover in aided hearing-impaired listeners
Background: At low presentation levels, a level increase typically improves audibility and thus speech intelligibility. At high presentation levels, a level increase can lead to poorer speech intelligibility. Termed rollover, this effect has been observed in listeners with normal and elevated audiometric thresholds. In a previous study, we found rollover at above-conversational levels in normal-hearing listeners in background noise when tested with context-free but not context-rich speech materials. We therefore concluded that semantic context information can mask rollover because of compensatory top-down mechanisms. However, other studies observed rollover at conversational levels in aided hearing-impaired listeners who were tested with context-rich speech materials only. In view of these discrepant findings, the current study investigated the influence of semantic context information on rollover at conversational and above-conversational presentation levels in aided hearing-impaired listeners.
Methods: Listeners with mild-to-moderate sensorineural hearing losses participated in speech intelligibility measurements in stationary speech-shaped noise. To compensate for their elevated hearing thresholds, individual linear amplification was provided by means of a wearable hearing-aid simulator. The speech materials included context-free sentences from the Danish DAT corpus and context-rich sentences from the Danish HINT corpus. The speech signals were presented at three levels: 65, 75, and 85 dB SPL.
Results: Initial results from this ongoing study indicate rollover with both types of speech materials. Performance decreases seem to occur at moderate levels (65 vs. 75 dB SPL) and particularly higher levels (75 vs. 85 dB SPL). Overall, this would seem to suggest that hearing-impaired listeners are susceptible to distortions arising in the auditory system at conversational and above-conversational levels, regardless of the availability of semantic context information.