作者: Guha Balakrishnan , Fredo Durand , John Guttag
关键词: Pulse (music) 、 Feature extraction 、 Object detection 、 Artificial intelligence 、 Optical flow 、 Motion estimation 、 Beat (music) 、 Heart rate 、 Computer vision 、 Heart rate variability 、 Electrocardiography 、 Computer science 、 Principal component analysis
摘要: We extract heart rate and beat lengths from videos by measuring subtle head motion caused the Newtonian reaction to influx of blood at each beat. Our method tracks features on performs principal component analysis (PCA) decompose their trajectories into a set motions. It then chooses that best corresponds heartbeats based its temporal frequency spectrum. Finally, we analyze projected this identify peaks trajectories, which correspond heartbeats. When evaluated 18 subjects, our approach reported rates nearly identical an electrocardiogram device. Additionally were able capture clinically relevant information about variability.