Adapting remote photoplethysmography for Indonesian subjects: an examination of diverse rPPG techniques

Abstract

Vital sign measurements are essential in intensive care patients, such as in the ICU or emergency department, and also for newborns or prenatal babies. The duty nurse usually monitors these vital signs by manually writing down the patient's condition on a large piece of paper in front of the patient's room. The lack of nurses can hinder the process of monitoring patient vital signs. However, since the COVID-19 pandemic, people have limited contact with their surroundings, making measuring vital signs with contact uncomfortable and unhygienic. The typical non-contact method for measuring heart rate is the remote photoplethysmography (rPPG) technique. In this study, we proposed to assess the performance of various rPPG algorithms on the Indonesian subjects dataset. The algorithms used are CHROM, GREEN, ICA, LGI, PBV, PCA, and POS on 70 pieces of data. Based on the test results with three types of evaluation metrics, namely MAE (Mean Absolute Error), RMSE (Root Mean Square Error), and Bland Altman, it is found that the measurement results with the best performance POS algorithm with a low prediction error rate with the resulting MAE value of 2.59 and RMSE of 4.65.