HrvMarquette: Heart Rate Variability Estimation

Title: hrvMarquette: Heart Rate Variability Estimation using Face Videos for Athletic Performance Evaluation

Mentor: Dr. Sheikh Iqbal Ahamed

Approach: Use remote photo plethysmography signals extracted from human face videos and analyze them using different algorithms to identify the peaks of the signal and thus measure their heart rate variability. The UBFC-rPPG dataset will be used for training the model and validating the method. The tool will be used to monitor and evaluate athletic performances and health conditions of athletic professionals. Summary: Heart rate variability (HRV) is a reliable parameter to identify several vital health conditions including an athlete’s stress level, fatigue and emotional state. Measuring HRV of humans using a non-contact method is a particularly challenging problem which has tremendous potential of being applied in professional athletic environments for performance and wellness monitoring. Accurate and real-time monitoring of an athlete’s stress level and health condition while training can provide important information to assist training and performance evaluation of athletes. The goals of this project are to implement an adaptive band pass filtering algorithm to accurately detect heart rate peaks using remote photo plethysmography (rPPG) methods, calculate heart rate variability, and finally train and validate the algorithm using UBFC-rPPG database. The algorithm will first detect the face and the skin to identify the region of interest. Then the mean value of RGB channels are calculated and then the pulse signal will be extracted. Finally, HRV will be calculated from the difference of the consequent RR intervals. Then, the Univ. Bourgogne Franche-Comt Remote Photo Plethysmography (UBFC-rPPG) data set will be used to further optimize the model and validate the results. This database contains human face videos obtained using Logitech C920 HD Pro webcam at 30fps and a resolution of 640x480 along with the HR and HRV values associated with it. The algorithm can be tested and trained using those values.

Student Research Activities: The REU fellows will perform the following major tasks:

• Detect face and skin to select the region of interest.
• Compute mean value of RGB channel pixels and perform amplitude selective filtering.
• Apply rPPG extraction using Plane orthogonal to skin tone algorithm.
• Complete signal processing by CWT and detect ipeak.
• Find RR interval and compute HRV.
• Evaluate the results with the stakeholders of the project.

Student Background: Students need to have moderate level programming skills in any high level programming language.