Heart rate estimation from facial videos with motion interference using T-SNE-based signal separation

• DOI: 10.1364/boe.457774
• Publication Date: 2022-07-18T15:30:08Z
• Authors: Hequn Wang; Xuezhi Yang; Xuenan Liu; Dingliang Wang
• Keywords: Photoplethysmogram; Robustness; SIGNAL (programming language)

Remote photoplethysmography (RPPG) can detect heart rate from facial videos in a non-contact way. However, head movement often affects its performance the real world. In this paper, novel anti-motion interference method named T-SNE-based signal separation (TSS) is proposed to solve problem. TSS first decomposes observed color traces into pulse-related vectors and noise using T-SNE algorithm. Then, it selects vector with most significant spectral peak as pulse for measurement. The tested on self-collected dataset (17 males 8 females) two public datasets (UBFC-RPPG VIPL-HR). Experimental results show that outperforms state-of-the-art methods, especially containing movements, improving Pearson correlation coefficient by 5% compared best contrasting method. To summarize, work significantly strengthens motion robustness of RPPG, which makes substantial contribution development video-based detection.