Facial Video based Heart Rate Estimation for Physical Exercise

Abstract

Heart rate estimation from facial videos is useful in applications such as telemedicine, public health monitoring, driver assessment, stress management and affective computing. Various studies have been done on evaluating remote photoplethysmography (rPPG) signals for subjects under different facial expressions to predict emotions. This paper proposed an analysis of heart rate measures from facial videos in the presence of heart rate variations for fitness applications. It is important to retrieve the health status of exercise and an optimized training program can be customized according to the preference physiological parameters. The state-of-the-art algorithm is applied to the raw RGB signals using Independent Component Analysis (ICA) method. The time-frequency domain of Fourier Transform is constructed to form PPG signals and estimate the heart rate. The analysis was carried out and validated using self-collected dataset using heart rate monitoring system prototype which developed using a pulse sensor as an input and Arduino microcontroller. The experimental results show that the state-of-the-art algorithm has an obvious low error index to proof efficiency and accuracy in various conditions of subjects with faster heartbeat after performing several physical exercises.