A Fast FPGA Hardware Accelerator for Remote Heart Rate Detection Based on RGB Vision

Area of Science:

• Biomedical Engineering
• Computer Engineering
• Signal Processing

Background:

• Remote photoplethysmography (rPPG) non-invasively measures physiological signals by analyzing skin color changes.
• Existing software-based rPPG methods for heart rate (HR) estimation can be computationally intensive.
• There is a need for faster, hardware-based solutions for real-time HR monitoring in applications like athlete health and driver safety.

Purpose of the Study:

• To develop and evaluate a fast hardware accelerator for HR estimation using rPPG technology.
• To achieve significantly faster calculation speeds compared to software implementations.
• To validate the system's accuracy and performance against a commercial device.

Main Methods:

• A hardware accelerator was designed using Verilog HDL and implemented on an Altera DE10-Standard FPGA board.
• Independent Component Analysis (ICA) was employed to extract blood volume pulse signals from RGB camera video data.
• Image capture, HR algorithm processing, and image display functionalities were integrated into the FPGA system.

Main Results:

• The hardware accelerator achieved high accuracy in HR estimation, with results of -0.76 ± 5.09 bpm (16s) and -0.70 ± 8.71 bpm (8s).
• The system demonstrated superior performance compared to prior works in terms of combined computation time and accuracy.
• Experimental validation was performed using a TRDB-D5M camera and an Omron HEM-6111 as the gold standard.

Conclusions:

• The proposed FPGA-based hardware accelerator offers a fast and accurate solution for HR estimation via rPPG.
• This technology has potential applications in real-time health monitoring for athletes and drowsiness detection for drivers.
• The hardware implementation significantly improves upon the computational speed of software-based rPPG methods.