Detail-Preserving Signal Fitting for Pulse Wave Detection from Smartphone-Based Fingertip Videos

Xuenan Liu,Xuezhi Yang,Jing Jin,Shuai Fang

DOI: https://doi.org/10.1109/icip.2019.8803243

2019-09-01

Abstract:With integrated LED lamps and cameras, smartphones are capable of extracting pulse waves from fingertip videos using image photoplethysmography technique. This paper presents a novel method for detecting pulse waves based on smartphone videos, with a focus on preserving vital details in pulse waves (such as limbs and dicrotic waves). Chrominance features are first extracted from videos to obtain a raw pulse wave, whose primary frequency is then used to build its fundamental pulsatile wave. After extracting pulse details from the raw pulse wave, a smooth pulse wave with clear details can be reconstructed by combining the fundamental pulsatile wave with pulse details. Experiments are conducted on a dataset involving 40 videos from 10 subjects under ambient lighting environments. Results have demonstrated the proposed method outperforms state-of-the-art ones in both pulse rate detection and pulse detail preservation especially in cases of motion artifacts.