Project outline
Evaluating the effectiveness of hearing aid algorithms can be a challenging task as evaluating different environments and conversational scenarios involves a large amount of preparation and effort on the behalf of both researchers and their participants.
Using virtual reality, we intend to produce a synthetic environment that allows for a balance between reproducibility and controllable variation. Participants wearing a virtual reality headset can participate as a listener in a virtual conversational environment, where the audio they hear is resynthesized in real-time from a ground truth dataset in response to their movements and behaviour within the virtual environment. By applying variations in the algorithms and parameters used for target speaker extraction and audio processing, we can make more reproducible assessments of both the impact of these variations in the experience of individual users, and evaluate which aspects of the transformations vary the most in accordance with user preference.
We expect that the insight gained from this process could better inform hearing aid manufacturers as to what customizations or fitting processes they could offer to their users to better align the hardware and software with their needs.
Project Partners
This project is hosted at the University of Sheffield and supported by the RNID.
Student
Thomas Stolarski