| Project Detail |
Over 400 million people worldwide suffer from disabling hearing loss. While the majority can be helped with a hearing aid, users still experience major difficulties in situations where multiple people talk simultaneously, leading to social isolation. Although algorithms exist to extract a single speaker from such a speech mixture, the current bottleneck is that a hearing aid does not know which of these speakers the user aims to attend.
Recent research has shown that this auditory attention can be decoded from the brain via electroencephalography (EEG). However, a practical realization in hearing aids is hindered by current EEG form factors: headsets are too bulky, while miniaturized sensors lack a wide scalp coverage. Furthermore, conventional EEG requires applying a gel on the skin, which dries out over time.
To address these issues, we propose a novel modular approach using 4 wireless EEG sensor patches strategically positioned on hairless scalp locations. This offers several advantages, such as (1) increased scalp coverage for improved decoding accuracy while maintaining comfort and discreteness due to a far-driven miniaturization and absence of wires; (2) the use of dry adhesive patches with skin-conforming microneedle electrodes for high-quality EEG recording without gel; (3) efficient algorithms that process signals and exchange compressed data wirelessly over short distances with minimal energy usage, thereby maximizing battery life.
In this project we will develop an operational prototype of this system drawing from our recent ERC-originated innovations, and validate it on hearing-impaired subjects. |
