Heart rate measurement is a key requirement for real-time health monitoring, particularly for elderly care. Traditional methods rely on contact sensing mechanisms, such as medical devices or wearable sensors like the Apple Watch. This paper presents a system for non-contact, real-time heart rate measurement using commodity cameras, like those embedded in laptops, where an innovative algorithm captures relevant signals to compute heart rate in real-life environments.
The heart rate computation (HRC) process consists of four major steps: (a) identifying the frames per second of the camera in use, e.g., 30 frames per second; (b) face detection (FD) using a deep learning method with a shape predictor of 68 facial landmarks; (c) applying a time sliding window (TSW) algorithm to de-noise the signal; (d) computing heart rate based on identified signal periodicity.
We tested and analyzed the developed prototypes against heart rate results from the Apple Watch, checking the difference range across multiple rounds and computing the mean difference for the same person's heart rate at the same time. Future work will involve further tuning and optimization of the methods, deploying the system as a personal AI agent for health monitoring.
Blogger's Review: This paper showcases an innovative non-contact heart rate measurement technology that leverages image processing and deep learning, offering substantial potential for application in elderly health monitoring. The approach enhances measurement convenience and provides a fresh perspective for future personal health management.