Tracking breathing, sleep Position, apnea and snoring, with vibrating alerts and data graphs generation
'SleepMonitoring' is a small, battery-powered, open‑source sleep monitoring device based on a single inertial sensor (IMU). It extracts breathing rate, sleep position, movements, suspected sleep apnea events by measuring low‑frequency micro‑movements of the human body. It also measures audio noises to detect snoring. Alerts can be triggered in case of position on the back that favors apneas and snoring. Snoring volume can also trigger vibration alerts.
The project is designed for experimentation. The software is Arduino based, in C. The user interface generating graphics is written in Python.
Quick start guide, with how to test with the user commands to interface with the board here
- Breathing rate estimation (breaths per minute)
- Detection of suspected sleep apnea events (breathing pauses)
- Sleep position detection (back, sides, stomach, upright)
- Movements detection/counting and discrimination from breathing
- Optional vibration feedback to discourage sleeping on the back, as people tend to have apneas or/and snore only when on the back
- Optional vibrations when snoring event is captured by microphone.
- Local data logging to internal flash memory (once per minute)
- Data captured can be displayed on computer by sending commands to the board from a user interface, through the USB connection.
- Circular flash memory management to maximize flash lifetime
- MCU: Seeed Studio XIAO nRF52840 Sense (Sense specific version is required only if the snoring detection is used, as it includes a PDM mic to capture sound)
- IMU (integrated into the XIAO nRF52840: LSM6DS3 (3‑axis accelerometer + 3‑axis gyroscope)
- Status RGB LED
- Push button
- Vibration motor
- Battery powered
- Plastic enclosure (small electric box)
The hardware is intentionally minimal to keep power consumption low and ensure easy reproduction.
Each breathing cycle produces small, periodic mechanical movements of the torso. These movements are captured by the accelerometer at very low frequencies (typically below 1 Hz).
The firmware:
- samples acceleration data continuously
- filters out high‑frequency noise and gross movements
- extracts low‑frequency oscillations
- detects breathing cycles and computes respiration rate
Data from accelerometer, with derivate computation to detect moves:
Breathing is distinguished from voluntary movements using:
- signal amplitude
- signal duration
- temporal regularity
Short, irregular, high‑amplitude signals are classified as movements. Low‑amplitude, periodic signals are classified as breathing.
Sleep position is determined using the direction of the gravity vector measured by the accelerometer. The firmware classifies the dominant orientation into predefined positions (back, left side, right side, stomach, upright).
Suspected apnea events are detected when:
- the breathing signal disappears or drops below a defined threshold
- for longer than a configurable duration
These events are logged for later offline analysis. The device does not provide medical diagnosis.
If enabled, a vibration motor is triggered when the subject remains on their back for too long. This gentle feedback encourages a spontaneous change to a side position, a common technique used in positional therapy.
Its purpose is to detect sustained snoring sounds during sleep and trigger a vibration alert that encourages the sleeper to change position or briefly wake up, reducing or stopping snoring episodes. The system continuously monitors ambient sound using the onboard PDM microphone. Instead of reacting to raw sound peaks, it estimates a short-term sound energy (volume) and compares it against adaptive thresholds that follow the background noise level.
To avoid false detections caused by noise of movements, audio detection is gated by motion analysis using inertial sensor data. Only sound events that occur while the body is relatively still are considered valid snoring candidates.
When repeated snoring events occur within a defined time window, the system triggers a vibration pattern via a motor to prompt a physical response from the sleeper. More details on snoring here:spec_snoring.md
-
Data is recorded once per minute
-
Logged data includes:
- breathing rate (average per minute and max per minute)
- sleep position: LEFT, RIGHT, LEFT, STOMAC, STANDING
- movements count
- Snorings count
To preserve flash memory lifetime, the firmware implements:
- circular flash storage
- wear‑optimized erase/write strategy
- no repeated erase of the same flash sector
This allows long‑term use without premature flash degradation.
Recorded data can be read directly using the Arduino IDE or the specific monitoring tool that can send debugging commands to the board and build monitoring graphs from the data retrieved on the board.
Example of graph built from the captured data:
The device is designed for overnight operation on battery power. Low‑power modes of both the MCU and IMU are used whenever possible to minimize consumption.
There is still work to be done in this domain to further reduce power consumption.
Functional and tested, certainly still some bugs or corner cases
This project is not a medical device and is not intended for diagnosis or treatment of any medical condition. It is provided for experimentation, learning, and personal data exploration only.
This project is released under an open‑source license. See the LICENSE file for details.
A detailed technical article describing the design and results is available on Hackaday.io.
Contributions, discussions, and improvements are welcome. Feel free to open issues or pull requests.