运动鲁棒的人脸视频心率检测算法  

Motion-robust Heart Rate Measurement Algorithm for Face Videos

作　　者：陈龙保 CHEN Longbao(Luoyang Polytechnic,Luoyang Henan 471000)

机构地区：[1]洛阳职业技术学院,河南洛阳471000

出　　处：《软件》2025年第3期63-66,共4页Software

摘　　要：心率是人体重要的生理指标之一,实时检测心率有助于健康管理。近年来,基于人脸视频的心率检测技术备受关注,其具有无接触、卫生、方便等优点。然而,在头部运动的情况下,远程心率检测技术与接触式检测相比精度较低。针对此问题,本文提出一种运动鲁棒的人脸视频心率检测算法。首先,检测人脸关键点划分感兴趣区域,提取原始色度信号;其次,对色度信号进行运动补偿,消除运动引起的信号漂移,选取信噪比较高的信号组成原始信号矩阵;最后,利用矩阵分解提取心率信号并计算心率。实验结果表明,本文方法在运动情况下有较强的鲁棒性。Heart rate is one of the important physiological indicators of the human body,and real time heart rate measurement helps in health management.In recent years,heart rate measurement technology based on face video has attracted much attention,which has the advantages of contactless,hygienic and convenient.However,in the case of head movement,remote heart rate measurement technology has lower accuracy compared with contact measurement.To address this problem,we propose a motion robust heart rate measurement algorithm for face video.Firstly,the key points of the face are detected to divide the region of interest,and the original chrominance signal is extracted;Secondly,the chrominance signal is motion-compensated to eliminate motion-induced signal drift,and the signal with higher signal-to-noise ratio is selected to form the original signal matrix;Finally,the heart rate signal is extracted using matrix decomposition and the heart rate is calculated.The experimental results show that the method in this paper has strong robustness in the case of motion.

关 键 词：人脸视频 色度信号 矩阵分解 心率检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]