PulseLens

Inspiration

Essential diagnostic tools like pulse oximeters and bedside monitors remain inaccessible or prohibitively expensive for rural clinics and low-resource settings. We built PulseLens to remove hardware barriers: converting ordinary consumer webcams and smartphone cameras into contactless physiological monitoring tools using computer vision and photoplethysmography (rPPG).

What it does

PulseLens captures micro-color variations across facial skin tissue caused by pulsatile blood flow to compute vitals in real time:

• Contact-Free Pulse Extraction: Computes estimated heart rate (BPM) and heart-rate variability indicators without external sensors.
• Respiratory Rate Tracking: Monitors subtle thoracic and facial rhythmic motion to calculate breathing frequency.
• Triage Alerts: Automatically flags abnormal vital patterns (tachycardia, bradycardia) for immediate clinical attention.
• Localized Privacy: Processes all video frames on the edge/client device, ensuring patient biometric data never leaves the hardware.

How we built it

• Computer Vision: OpenCV and MediaPipe facial mesh tracking to isolate stable regions of interest (ROI) across forehead and cheek zones.
• Signal Processing Pipeline: Temporal bandpass filtering (0.7 Hz to 4.0 Hz / 42–240 BPM) combined with Fast Fourier Transform (FFT) peak frequency estimation.
• Application & Interface: Python and Streamlit dashboard delivering instant live-stream telemetry and diagnostic summaries.

Challenges we ran into

Mitigating optical noise and motion artifacts caused by slight head tilts and uneven ambient lighting. We solved this by implementing adaptive facial landmark anchoring and dynamic spatial RGB averaging across skin contours.

Accomplishments that we're proud of

• Eliminating reliance on expensive medical peripheral hardware.
• Maintaining sub-second latency and smooth FPS entirely within local edge compute.

What we learned

Applying mathematical signal decomposition directly to optical camera feeds, and designing digital health systems specifically resilient to real-world edge hardware constraints.

What's next for PulseLens

• Adding oxygen saturation (SpO2) and real-time stress index estimation.
• Packaging into a progressive web app (PWA) with offline health summary exports for rural field workers.

Built With

• computer-vision
• digital-health
• mediapipe
• numpy
• opencv
• python
• rppg
• scipy
• streamlit