Cor

Inspiration Modern workflows demand intense focus, but that focus often comes at the invisible cost of physiological stress. We noticed that during long, deep-work sessions—whether debugging complex code or grinding through extensive research and flashcard reviews—our bodies would enter a state of continuous sympathetic nervous system activation. Heart rates creep up, breathing becomes shallow, and tension builds without us even realizing it until the session ends. While wearables exist to track this, they require the user to actively wear another device and often just present raw, anxiety-inducing data. We wanted to build a frictionless, software-only solution that understands when you are stressed and actually steps in to help you fix it. What it does Cor is a passive, desktop-based stress monitor that utilizes your existing webcam to track your heart rate in real time without any wearable hardware. Operating seamlessly in the background, Cor functions in two primary states: The Passive State: It runs quietly in the system tray, periodically sampling your webcam feed to track baseline physiological metrics without distracting you or displaying a continuous video feed. The Active State (Intervention): When Cor detects a sustained, abnormal spike in your heart rate that correlates with stress rather than physical movement, it triggers a non-intrusive UI notification. If accepted, the app expands into a clean, rhythmic breathing interface designed to manually activate your parasympathetic nervous system and safely guide your heart rate back to a normal baseline. How we built it The core engine of Cor relies on remote photoplethysmography (rPPG). We built a computer vision pipeline using Python and OpenCV to isolate specific regions of interest on the user's face. By tracking the micro-fluctuations in skin color caused by the volumetric change of blood flow with each heartbeat, we are able to extract a raw pulse signal. We then process this raw optical data through signal filtering algorithms to separate the true heart rate from ambient noise. Finally, we took this backend algorithmic logic and wrapped it in a clean, modern UI architecture, managing the heavy computer vision tasks on background threads to ensure the frontend intervention graphics remain smooth and responsive. Challenges we ran into Extracting a clean, accurate rPPG signal from a standard webcam is incredibly complex. The biggest hurdles were environmental variables: handling dynamic lighting changes (like a monitor screen shifting from dark mode to a bright white webpage) and filtering out motion artifacts when the user shifts in their chair or types aggressively. Additionally, transitioning our raw OpenCV ⁠cv2.imshow⁠ debugging loops into a consumer-ready desktop application required careful thread management. We had to ensure that running a continuous background facial-tracking model didn't hog CPU resources or drain battery life, defeating the purpose of a background utility. Accomplishments that we're proud of We are incredibly proud of bridging the gap between clinical-grade biometric algorithms and intuitive user experience. It is one thing to get an accurate heart rate from a webcam feed in a controlled environment; it is entirely another to package that technology into a sleek, unobtrusive tool that feels like a natural extension of a user's daily workflow. Moving from raw data detection to active physiological de-escalation makes Cor a truly useful wellness tool. What we learned We learned a massive amount about optimizing computer vision pipelines for real-time, low-resource execution. Beyond the technical code, we learned a valuable lesson in UX design for health-tech: simply showing a user that their heart rate is spiking can actually induce more stress. We learned that actionable, guided de-escalation (like synchronized breathing UI) is vastly more effective than just displaying a dashboard of raw metrics. What's next for Cor Our immediate next steps involve refining the application's performance in extreme low-light environments, specifically catering to late-night coding or study sessions. We also plan to explore integrating the data securely with calendar and task-tracking systems, allowing users to retrospectively identify exactly which types of tasks or meetings consistently trigger their highest stress responses.

Built With

• opencv
• python