Abstract
Sleep quality strongly affects learning, but conventional sleep staging needs clinical equipment and expert review. We propose a collaborative-reasoning framework using locally deployed LLMs for EEG sleep staging on edge devices, preserving privacy while improving accuracy through multi-model cooperation.
Problem & Motivation
Sleep quality has deep effects on learning outcomes, but conventional sleep monitoring requires expensive clinical-grade equipment and expert interpretation. How can ordinary students obtain accurate sleep-stage analysis from affordable EEG devices? And given that sleep data is sensitive personal information, how do we avoid the privacy risk of cloud-side analysis?
Method
We propose a collaborative-reasoning framework that uses locally deployed LLMs to perform EEG sleep staging. The system runs on edge devices, avoiding cloud upload of sensitive physiological data, while multi-model collaborative reasoning lifts staging accuracy.
Findings
- Locally deployed LLMs can perform effective EEG sleep staging on edge devices.
- Collaborative-reasoning framework improves accuracy over single-model staging.
- Edge deployment preserves the privacy of users' physiological data.
Implications
Understanding student sleep state helps instructors better understand learner state. Edge-deployed AI analysis lets physiological data be applied at lower privacy cost, more accessibly — a technical foundation for the in-development Uedu Brain wearable.
Citation
BibTeX
@inproceedings{cheng2025eeg_sleep,
author = {Cheng-Lin Cheng and Ting-Chuan Lin and Chia-Kai Chang},
title = {Collaborative Reasoning Framework for Edge-Deployable {EEG} Sleep Staging via Local {LLM}},
booktitle = {Proc. IEEE Int. Conf. on Big Data Computing Service and Machine Learning Applications (BigDataService)},
pages = {108--112},
year = {2025},
month = jul,
doi = {10.1109/BigDataService65758.2025.00021},
}