Project DetailsThe project’s goal is to develop a dual-mode system for contactless vital-sign monitoring from 1440p facial video using remote photoplethysmography (rPPG). The prototype, implemented on an NVIDIA Jetson Orin Nano Super Developer Kit, records an uncompressed video stream to a microSD Express card while simultaneously running a downscaled, low-latency pipeline that provides live feedback.
Utilizing microSD Express over PCIe/NVMe enables sustained 30 fps uncompressed 1440p video capture—bandwidth that legacy microSD solutions cannot reliably support—thus allowing practical high-fidelity offline analysis. In evaluation against a Polar Verity Sense reference over a dataset of 10 recordings, the system achieved an MAE of 8.3 bpm in real-time mode and 6.6 bpm in offline processing. These results highlight the accuracy improvements made possible by preserving uncompressed high-resolution video for post-capture analysis.
By quantifying I/O throughput and overall system performance, and demonstrating a fully functioning prototype, the project illustrates the role of microSD Express as a key enabler for edge-based vision workloads in consumer telehealth and wellness applications.
