Improving Signal Quality for Remote Photoplethysmography by Suppression of Motion and Lighting Effects.

Remote photoplethysmography (rPPG) is a novel non-contact method for physiology measurement. This method shows limitations in signal quality due to the motion and lighting disturbances, which hinder its application in daily life. This study aimed to develop a novel anti-noise algorithm for the motion and lighting disturbances so as to improve the quality of rPPG signals. This study designs a light interference suppression module and a specular reflection interference suppression module to improve the rPPG system by analyzing the properties of the noise introduced by motion and light disturbances. The light interference suppression module uses log-domain filtering to demodulate and remove the lighting component. The specular reflection suppression module uses weighted least squares Kalman filter to dynamically suppress unpredictable motion artifacts. We used the PURE dataset to verify the proposed algorithm, and evaluated the signal quality by calculating the error in pulse rate estimation. For different unsupervised rPPG extraction algorithms, the results show that the interference suppression module effectively reduces the error of pulse rate estimation. The mean absolute error of the enhanced system employing the CHROM method is 0.6688 bpm, representing a 19.55% reduction compared to the basic system; the root mean square error is 1.2379 bpm, representing a 43.62% reduction; and the correlation improves from 0.9963 to 0.9990. The Bland-Altman plots show that our algorithm can effectively suppress abnormal outliers caused by noise. This study may offer novel insights into the suppression of noise in rPPG signals and will play a role in daily physiological measurement.Clinical Relevance- This study established a denoising algorithm to improve the robustness of rPPG-based non-contact physiological parameter detection in clinical and daily life.