🧬 biology

OSSMM: An Open-Source Sleep Monitor and Modulator

The paper introduces OSSMM, a low-cost, open-source wearable headband using repurposed fitness electrodes and an Android app to accurately classify four sleep stages and offer modulation capabilities, demonstrating that practical sleep monitoring is achievable with simplified, accessible hardware configurations.

Original authors: Jonny Giordano, Fergal Stapleton, Gabriel Palma, Barak A. Pearlmutter

Published 2026-05-19

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Jonny Giordano, Fergal Stapleton, Gabriel Palma, Barak A. Pearlmutter

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Imagine trying to understand the secret language of your brain while you sleep. For decades, scientists have needed a massive, expensive hospital setup with sticky wires and gels to translate that language. This paper introduces a new, open-source project called OSSMM (Open-Source Sleep Monitor and Modulator) that aims to translate that language using a simple, cheap headband you could build yourself for less than the price of a nice dinner (under €40).

Here is the story of how they did it, broken down into simple parts:

1. The Problem: Sleep Research is Too Expensive and Fussy

Think of traditional sleep studies (Polysomnography) like a high-end, custom-tailored suit. It's the "gold standard," but it's heavy, uncomfortable, costs thousands of dollars, and you can only wear it in a specific room with a technician watching you. Because it's so hard to get, researchers often have to choose between studying a few people very carefully or many people very poorly.

The authors wanted to build a "denim jacket" of sleep monitors: something affordable, reusable, easy to put on, and open for anyone to fix or improve.

2. The Solution: A DIY Sleep Headband

The team built a small headband using 3D printing and parts you can buy at any electronics hobby shop.

The Hardware: It's a lightweight headband (about the size of a large coin) that holds a tiny computer, a vibration motor (to potentially wake you up or change your sleep later), and sensors.
The "Secret Sauce" Electrodes: Instead of using expensive, disposable medical electrodes, they used conductive plastic strips taken from cheap fitness chest straps (the kind runners wear to track heart rate). They wiped these clean with a damp cloth and reused them every night.
The Setup: The headband connects wirelessly to a standard Android phone, which acts as the brain of the operation, storing the data and displaying the graphs.

3. The Experiment: One Person, 15 Nights

To test if this "denim jacket" could actually work, one of the researchers wore the headband for 15 nights.

The Rival: Every night, they also wore a high-tech, non-contact sleep monitor (the Somnofy) on their bedside table. This device is already proven to be pretty accurate compared to hospital tests.
The Goal: They wanted to see if the cheap headband could "speak the same language" as the expensive bedside monitor. Could the headband tell the difference between being awake, light sleep, deep sleep, and dreaming (REM)?

4. The Results: The Headband "Got It"

The team fed the data from the cheap headband into three different computer learning programs (think of them as three different students trying to solve a puzzle).

The Winner: One program (called Random Forest) got it right about 77% of the time.
The Comparison: This is a very strong result for such a simple device, especially since the computer wasn't given any special tricks or complex math to help it. It just looked at the raw signals and figured it out.
The "Aha!" Moment: The computer realized that the most important clues came from the brain waves picked up by the cheap fitness straps. Even though these straps weren't designed for brain waves, they managed to catch specific "brain signatures" (like sleep spindles, which are little bursts of brain activity during light sleep) that are usually only seen with expensive medical gear.

5. Two Big Surprises

The paper highlights two things that surprised the researchers:

The Cheap Straps Work: They found that the conductive plastic from a $5 fitness strap could pick up brain signals that looked exactly like the complex brain waves seen in hospitals. It's like finding out a paperclip can conduct electricity just as well as a copper wire for a specific job.
Fewer Wires Needed: Usually, to measure brain waves, you need a "ground" wire (like a safety anchor) to make the signal clear. This headband worked perfectly with just two wires on the forehead and no ground wire at all. It's like driving a car with only two wheels instead of four and still managing to stay on the road.

6. What This Means (and What It Doesn't)

What it means: The barrier to entry for sleep research has been lowered. Anyone with a 3D printer and basic electronics skills can now build a device that captures useful sleep data. The data is "open-source," meaning the blueprints are free for everyone to use, copy, and improve.
What it doesn't mean (yet): This was a test on one person. We don't know if it works for everyone (like children, the elderly, or people with different sleep disorders) until more people try it. Also, while it matched the bedside monitor well, it hasn't been compared directly to the "gold standard" hospital test yet.

Summary

The OSSMM project is like building a homemade telescope. For a long time, you needed a massive, expensive observatory to see the stars. This team built a small, cheap telescope out of household parts that can still see the stars clearly. They proved that you don't need a million-dollar lab to start understanding sleep; you just need a little creativity, some 3D printing, and a willingness to share the blueprints with the world.

Technical Summary: OSSMM: An Open-Source Sleep Monitor and Modulator

Problem Statement
Sleep research faces significant barriers regarding cost, accessibility, and ecological validity. Traditional polysomnography (PSG) is the gold standard but is expensive (averaging $1,575–$ 1,840), requires clinical facilities, and involves uncomfortable, restrictive equipment. Conversely, low-cost alternatives often suffer from poor accuracy or require complex setups involving conductive gels and disposable electrodes. Furthermore, active sleep modulation research, which requires real-time state detection to target specific neural signatures, has been limited by the lack of affordable, transparent, and easily modifiable hardware. There is a critical need for a platform that balances affordability with sufficient data quality to support sleep staging and modulation without the high overhead of traditional PSG.

Methodology
The authors developed the Open-Source Sleep Monitor and Modulator (OSSMM), a wearable headband system built from 3D-printed components and commercial-off-the-shelf (COTS) electronics, with a total material cost under €40.

Hardware Architecture: The device utilizes a Seeed Xiao nRF52840 Sense microcontroller to coordinate data acquisition from three sensor boards: an IMU for 6-DoF movement, a PPG sensor for pulse, and an AD8232 EKG board configured for EOG/EEG acquisition. Notably, the system employs two conductive thermoplastic polyurethane (CTPU) electrodes embedded in a standard fitness chest strap. These electrodes are placed frontally to capture a differential signal without a dedicated ground reference electrode. A vibration motor is included for potential sleep modulation. Data is transmitted via Bluetooth Low Energy (BLE) to a companion Android application.
Experimental Design: The study employed a single-participant (N=1) proof-of-concept design to validate system functionality and hardware reliability before broader deployment. Data was collected over 17 nights, with 15 qualifying nights yielding 15,285 30-second epochs.
Reference and Labeling: Sleep stage labels were provided by a validated non-contact sleep monitor, the Somnofy SM-100 (which shows $\kappa=0.63$ agreement with PSG). The dataset was split into 12 nights for training and 3 nights for testing.
Signal Processing and Classification: Raw data was digitized at 250 Hz. Features were extracted from the 30-second epochs, including 8 pulse features, 12 IMU movement features, and 22 EOG/EEG features (including power in canonical frequency bands: delta, theta, alpha, beta, and gamma). Three machine learning classifiers—Support Vector Machine (SVM), Random Forest (RF), and XGBoost (XGB)—were evaluated using Group K-Fold cross-validation to prevent data leakage between nights. Hyperparameters were manually searched to assess baseline signal quality rather than to maximize performance through extensive tuning.

Key Contributions

Open-Source Platform: The OSSMM provides a fully transparent, low-cost (<€40) hardware and software platform. All designs, code, and build instructions are publicly available, allowing researchers to assemble, modify, and replicate the system.
CTPU Electrode Utility: The study demonstrates that inexpensive, reusable CTPU electrodes from commercial fitness chest straps can capture differential signals with spectral properties consistent with canonical EEG phenomena, including sleep spindles (11–16 Hz).
Simplified Configuration: The system achieves sleep staging using only two frontal electrodes without a dedicated ground reference, challenging the necessity of complex multi-electrode setups for practical classification.
Modulation Capability: The inclusion of an onboard vibration motor enables the platform to serve as a modulator for future sleep intervention studies.

Results
The study evaluated the ability of OSSMM signals to classify four sleep stages: Wake, Light Sleep, Deep Sleep, and REM.

Performance: The Random Forest classifier achieved the best performance on the withheld test set, yielding a Macro F1-score of 0.770 and an accuracy of 0.776. XGBoost followed closely with an F1-score of 0.741, while SVM performed significantly lower (F1=0.598).
Feature Importance: Analysis of the best-performing models revealed that EOG/EEG-derived features dominated the top rankings. Specifically, alpha band power was the most important feature, followed by EOG interquartile range (IQR). Spectral features from the CTPU electrodes were the primary drivers of classification success.
Signal Characteristics: Visual inspection of the data confirmed the presence of sleep spindle-like signatures (approx. 12 Hz, 0.5–2 seconds) in the EOG/EEG channel during Light Sleep, consistent with known EEG phenomena.
Classification Patterns: Models showed a tendency to confuse adjacent sleep stages (e.g., Light vs. Deep) but rarely misclassified distant states (e.g., REM vs. Deep Sleep). Deep Sleep recall was lower than other stages, a common challenge in sleep staging.

Significance and Claims
The paper claims that the OSSMM platform successfully demonstrates that affordable, open-source hardware can capture biosignals sufficient for practical 4-stage sleep classification. The authors emphasize two technical findings:

Signal Quality: The differential signal captured from simple CTPU chest strap electrodes contains neurophysiologically consistent information (including sleep spindles) that drives classification, suggesting these electrodes are viable for EEG-related research.
Simplicity: Accurate sleep staging is achievable with a simplified configuration using only two frontal electrodes and no ground reference, potentially lowering the barrier to entry for sleep research.

The authors maintain a modest stance, noting that the study is a technical validation against a reference device (Somnofy) rather than a gold-standard PSG comparison. They acknowledge limitations regarding the single-participant design and the need for future validation against PSG to confirm the cortical origin of the signals and generalizability to broader populations. However, they conclude that the OSSMM offers a viable, low-cost alternative that does not significantly sacrifice classification performance, thereby facilitating more accessible and transparent sleep research.