Optimal Digital Filter Selection for Remote Photoplethysmography (rPPG) Signal Conditioning.

Abstract

Remote photoplethysmography (rPPG) using camera-based imaging has shown excellent potential recently in vital signs monitoring due to its contactless nature. However, the optimum filter selection for pre-processing rPPG data in signal conditioning is still not straightforward. The best algorithm selection improves the signal-to-noise ratio (SNR) and improves the accuracy of the recognition and classification of vital signs. We recorded more than 300 temporal rPPG recordings, where the noise is mainly not motion-induced. Then, we investigated the best digital filter in pre-processing temporal rPPG data and compare the performances of ten different filters with ten orders each (i.e., total 100 filters). The performances are assessed using a signal quality metric on three levels as the quality of the raw signals was classified under three categories; Q1 being the best Q3 being the worst. The results are presented in SNR scores, which show that the Chebyshev II orders of 2nd, 4th, and 6th perform the best for denoising rPPG signals.