Improving the Accuracy of Camera-Based Heart Rate Measurement.

Area of Science:

• Computer Vision
• Biomedical Imaging
• Signal Processing

Background:

• Non-contact heart rate (HR) monitoring using visible light cameras shows promise for clinical and home healthcare.
• Accuracy is compromised by CMOS sensor rolling shutters and image acquisition time variations due to background processes.

Purpose of the Study:

• To analyze and mitigate phase errors from CMOS sensors and system errors from clock jitters in camera-based HR monitoring.
• To propose and validate novel methods for enhancing the accuracy of non-contact HR detection.

Main Methods:

• Amplitude-frequency superposition method to eliminate phase errors by analyzing signal amplitude in the frequency domain.
• Cubic spline interpolation reconstruction method based on actual schedules to correct non-uniform sampling caused by system clock jitters.

Main Results:

• Amplitude-frequency superposition improved signal amplitude by 4.58% in simulations and 33.5% in real measurements.
• Cubic spline interpolation reduced root mean square error by over 30% in simulations and improved HR accuracy by approximately 40% in real measurements.

Conclusions:

• The proposed methods effectively eliminate camera phase errors and reduce system clock jitter-induced random errors.
• These techniques offer significant improvements for dynamic signal detection in machine vision, particularly for non-contact HR monitoring.