EEG sleep band
A forehead EEG headband that plays short sounds in time with the brain's slow oscillations during sleep. Software first: the timing problem is solved and evaluated on recorded sleep data before any hardware is built.
- Sole developer: signal processing, evaluation, hardware design
- Neurotech
- ML
Problem
Deep sleep is dominated by slow oscillations, large waves of about one per second. Studies of closed-loop auditory stimulation have shown that short sounds played at the right moment of such a wave can enhance it, and that the same sounds at the wrong moment can suppress it. The effect depends on timing, and that is the hard part: the wave is irregular, it has to be tracked in real time from past samples only, and every device adds latency between deciding to play a sound and the sound reaching the ear.
Most systems that do this well are lab setups or closed products. The goal here is a comfortable forehead device whose timing can be measured and reported.
Approach
The project is split so that the risky part is answered first:
- Software: an offline evaluation on public sleep recordings, so that timing methods can be compared before any hardware exists.
- Hardware: a forehead EEG front end based on the ADS1299, first on development boards, then on an own PCB.
- Closed loop: overnight recordings, first without stimulation, then with it.
TODO: add method details once the disclosure question is settled.
Results
- TODO: timing accuracy metric
- —
- TODO: system latency
- —
- TODO: recording quality metric
- —
No results yet. The devlog below follows the project as it goes.
My role
Sole developer: signal processing, evaluation, firmware and hardware design. TODO: name advisors if any.
What I’d do differently
Too early to say. TODO: fill in after the first phase.
Devlog
Devlog #0: Why I'm building an EEG sleep band
Placeholder post that shows the note format: why this project, what comes first, and how progress will be logged.