Closed-loop neurostimulation system for sleep onset optimization, adapting stroboscopic light and generative audio to live EEG. Part of BioListener, an open-source ecosystem designed to revolutionize biosensing with portable, affordable, multi-channel biosensing boards.
Research stage. BioListener Sleep is researched in simulation and on an integration test harness; sessions with real subjects are a feasibility check only. It is research software, not a medical device, and makes no claim that it helps anyone fall asleep.

How it works
Most stimulation devices play a fixed program and read nothing from the person. BioListener Sleep measures the person while it stimulates, and keeps adjusting.
The BioListener board records brain activity (EEG) from passive electrodes on the scalp.
A controller trained with deep reinforcement learning chooses the next stimulation setting.
A separate safety supervisor clamps every requested action before it reaches the hardware.
The command drives a stroboscopic LED mask and an audio engine. Then the loop repeats.
The research
If a learned controller drives stimulation from live EEG, what should it be rewarded for? The project studies this reward design.
The controller learns in a simulated model of brain cortex, cross-validated against The Virtual Brain, a standard library in computational neuroscience.
The full loop has run with a person on the BioListener board and the LED mask, monitored from a browser console with a panic stop.
Questions, collaboration, or contributions: write to us, or open an issue or pull request in the relevant repository.