Investigating the Effectiveness of Filter in rPPG Methods Across Various Skin Tones

Heart rate is a crucial indicator of physical health, measured manually or via photoplethysmography (PPG) devices. This study assesses the performance of Elliptic, Chebyshev1, and Butterworth filters in processing remote photoplethysmography (rPPG) signals, focusing on noise reduction and accuracy with varying skin tones. Data from 19 subjects aged 20 to 32 years, from diverse ethnic backgrounds, are collected using video recordings and fingertip PPG sensors. Video data are preprocessed to isolate the face region, and rPPG signals are extracted and post-processed for amplitude normalization and frequency filtering. Results indicate that the Elliptic filter achieves the lowest root mean square error (RMSE) of 1.751 and the highest signal-to-noise ratio (SNR) of 5.536. The Chebyshev1 filter has an RMSE of 1.768 and an SNR of 3.066, while the Butterworth filter shows the highest RMSE of 3.411 and an SNR of 1.699. Among rPPG methods, ICA and PCA offer the best accuracy.