ESP32 Radar Thermal Infant Apnea Monitor Science Fair

By /

ESP32 Radar Thermal Infant Apnea Monitor Science Fair

ISEF Category: Biomedical Engineering

Ready to Turn This Idea Into a Real Project?

This guide was put together with the help of AI research tools to give you a solid starting point.But a competitive science fair project lives in the details: refining your research question, fine-tuning your variables, analyzing your data, and presenting your findings like a seasoned scientist.

For next steps tailored to your interests, skill level, and timeline, work one-on-one with a MehtA+ mentor. Learn more about MehtA+ Science & Engineering Research Mentorship →

Subcategory: Biomedical Devices  ·  Difficulty: Advanced  ·  Setup: Home Setup  ·  Time: 1 to 2 Months

The Hook

Sudden infant death syndrome (SIDS) still claims thousands of babies each year. Most consumer monitors require something stuck to the baby. A $30 board with radar and a low-resolution thermal camera detects breathing cessation across a crib without touching the infant at all. Tiny ML on the ESP32 keeps everything private and on-device.

What Is It?

HLK-LD2410 is a 24 GHz radar module costing about $5. It detects micro-motion such as chest rise. MLX90640 is a 32x24 pixel thermal sensor that captures heat patterns including breath plumes.

Fusing the two streams improves robustness. Radar handles motion, thermal handles temperature signatures, and an anomaly model spots breathing cessation patterns even when one signal is noisy.

Public neonatal apnea datasets like PhysioNet's Apnea-ECG provide labeled breathing patterns. Even though they are not radar/thermal, they let you tune the anomaly thresholds against real apnea episodes.

Why This Is a Good Topic

Non-contact infant monitoring is an active product space and a legitimate engineering problem. Sensors are cheap and the safety stakes give the project weight. You will learn sensor fusion, anomaly detection, and embedded ML deployment.

Research Questions

  • How does sensor placement above the crib change detection sensitivity?
  • What is the effect of room temperature on thermal signal?
  • Does fused radar plus thermal outperform either alone?
  • To what extent does ambient motion confound the radar?
  • Which model architecture fits inside the ESP32 memory budget?
  • How does sleeping position affect detection?
  • What is the effect of false-alarm threshold on parental confidence?

Basic Materials

  • HLK-LD2410 radar module.
  • MLX90640 thermal sensor.
  • ESP32 development board.
  • LiPo battery.
  • Mount and tripod for above-crib placement.
  • Test mannequin or breathing simulator.
  • Informed-consent form for any human pilot.

Advanced Materials

  • Higher-resolution thermal camera (FLIR Lepton).
  • Clinical apnea simulator.
  • IRB approval for infant data.
  • Hospital-collaboration mentor.

Software & Tools

  • TensorFlow Lite Micro: Deploys the anomaly model on-device.
  • Python (NumPy and PyTorch): Trains the model.
  • PlatformIO: Builds firmware.
  • PhysioNet datasets: Provides labels for model training.

Experiment Steps

  1. Lock the sensor mount geometry.
  2. Decide signal preprocessing (filter band, frame rate) and lock it.
  3. Train the anomaly model on PhysioNet and synthetic data.
  4. Run bench tests with a breathing simulator to bracket detection limits.
  5. Plan ethical pilot only with adult volunteers or simulator data.
  6. Compare detection sensitivity and false-alarm rate across configurations.

Common Pitfalls

  • Treating an unvalidated prototype as a medical device.
  • Mixing training data from very different sensors without domain adaptation.
  • Skipping ambient motion controls (fans, pets).
  • Reporting only sensitivity and hiding false-alarm rate.
  • Forgetting battery-life measurements at the chosen sample rate.

What Makes This Competitive

A competitive entry trains the anomaly model with cross-validation, compares fused vs. single-sensor baselines, and reports false-alarm rate alongside sensitivity. On-device latency, power consumption, and a safety-failsafe analysis push the design beyond a prototype.

Project Variations

  • Replace radar with a UWB module and compare detection range.
  • Add a piezo microphone for cry detection.
  • Test the device on adult volunteers performing controlled breath holds.

Learn More

  • HLK-LD2410 datasheet: Free official documentation.
  • MLX90640 application notes: Open documentation.
  • PhysioNet Apnea-ECG database: Free with documentation.
  • PubMed: Search non-contact apnea monitoring reviews.
  • TensorFlow Lite Micro guides: Free deployment tutorials.
Shopping Cart