Assessing hearing device benefit using virtual sound environments

The investigation into people’s ability to understand speech in noisy everyday situations, particularly those affected by hearing loss, constitutes an important area of hearing research. Hearing devices, such as hearing aids, attempt to restore a hearing-impaired person’s real-world hearing ability. However, many psychoacoustic tests currently in use to evaluate speech intelligibility and improve hearing aid performance do not take the acoustic properties of complex real-world sound scenes into account, instead relying on artificial target speech and background noise signals presented over headphones or small sets of loudspeakers. While such laboratory settings provide highly controlled and reliable results, they typically do not capture acoustic characteristics of real-world environments such as reverberation and moving sound sources, and they do not fully reflect how people experience their real-world auditory reality.
This thesis aimed to bridge the gap between laboratory-based hearing tests and real-world listening by evaluating hearing using loudspeaker-based virtual sound environments (VSEs). Such a VSE is reproduced inside a spherical array of loudspeakers, capable of presenting spatialized sound fields to a listener positioned in the center with a high level of physical accuracy. By employing VSEs in combination with spatially recorded real-world noise signals and spatialized target speech, acoustically realistic speech intelligibility tasks were designed and implemented. This included the development of a method for in-situ, realistic conversational signal-to-noise ratio estimation, intended to characterize a talker’s real-world speech levels. Measured speech reception thresholds (SRTs) were shown to be elevated for realistic VSE conditions compared to more artificial headphone and spatialized artificial noise conditions for both normal-hearing and hearing-impaired listeners. However, the hearingimpaired Furthermore, it was shown that hearing aid dynamic range compression benefited speech intelligibility more in the realistic VSE condition compared to more artificial conditions, likely as a consequence of the acoustic properties of the speech and noise signals and their effect on the hearing aid signal processing. Finally, a method for guided ecological momentary assessment (EMA) was conceived to evaluate subjective, listener-reported hearing ability in a way that would reduce the data variability found in conventional EMA. The proposed method was shown to result in consistent ratings of hearing ability across a group of normal-hearing participants. The ratings were found to be reproducible inside acoustically matched, realistic VSEs. Overall, this thesis showed the ability of VSE-based laboratory environments to provide increased acoustic realism in psychoacoustic listening tasks, rendering more ecologically valid results, for both normal-hearing and hearing-impaired individuals. The development of increasingly realistic VSE-based hearing and hearing aid evaluation tests has the potential to increase the benefit hearing devices provide to users in their everyday life. listeners’ SRTs increased more between the artificial conditions and the realistic VSE condition than those of the normal-hearing listeners. Speech recognition scores obtained at the normal-hearing conversational signal-tonoise ratio provided percentage-correct scores relating speech intelligibility performance to communication ability in the real world.
Furthermore, it was shown that hearing aid dynamic range compression benefited speech intelligibility more in the realistic VSE condition compared to more artificial conditions, likely as a consequence of the acoustic properties of the speech and noise signals and their effect on the hearing aid signal processing. Finally, a method for guided ecological momentary assessment (EMA) was conceived to evaluate subjective, listener-reported hearing ability in a way that would reduce the data variability found in conventional EMA. The proposed method was shown to result in consistent ratings of hearing ability across a group of normal-hearing participants. The ratings were found to be reproducible inside acoustically matched, realistic VSEs.
Overall, this thesis showed the ability of VSE-based laboratory environments to provide increased acoustic realism in psychoacoustic listening tasks, rendering more ecologically valid results, for both normal-hearing and hearing-impaired individuals. The development of increasingly realistic VSE-based hearing and hearing aid evaluation tests has the potential to increase the benefit hearing devices provide to users in their everyday life.