基于交叉滑动窗口模型算法的非接触心率测量方法  

Non-contact heart rate measurement method based on cross sliding window model algorithm

作　　者：汤春明[1] 张凯凯 王慧泉[2] 田磊[2] 于翔 TANG Chun-ming;ZHANG Kai-kai;WANG Hui-quan;TIAN Lei;YU Xiang(School of Electronics and Information Engineering,Tiangong University,Tianjin 300387,China;School of Life Sciences,Tiangong University,Tianjin 300387,China;Center for Engineering Practice and Training,Tiangong University,Tianjin 300387,China)

机构地区：[1]天津工业大学电子与信息工程学院,天津300387 [2]天津工业大学生命科学学院,天津300387 [3]天津工业大学工程教学实习训练中心,天津300387

出　　处：《天津工业大学学报》2020年第2期54-60,共7页Journal of Tiangong University

基　　金：武汉工程大学协作项目(2301/028663)。

摘　　要：为解决传统接触式心率测量系统设备昂贵、操作复杂的问题,在远程光电容积描记法(remote photoplethysmography,r PPG)测量心率的基础上,提出一种基于交叉滑动窗口模型(cross sliding window model,CSWM)算法的非接触心率测量方法。首先调用人脸检测模型SSD300检测视频中的人脸,然后使用Otsu算法对人脸皮肤区域进行分割,并用形态学方法腐蚀皮肤之间的眼睛、嘴巴区域,同时用KCF算法跟踪人脸,并将人脸区域均分为2×2的大小,然后使用CHROM算法提取rPPG波形,最后运用CSWM方法对rPPG波形进行处理,得到实时变化的心率。实验结果表明:该方法计算出的心率与同时用脉搏血氧仪测出的心率之间的平均准确率为97.9217%,Bland-Altman一致性系数达0.962302,完全满足临床要求。In order to solve the problem of expensive equipment and complicated operation in the traditional contact measurement system of heart rate,based on the remote photoplethysmography(rPPG)measurement of heart rate,a non-contact measurement of heart rate based on cross sliding window model(CSWM)algorithm is proposed.Firstly,the face detection model SSD300 is called for to detect the face in the video,and then the Otsu algorithm is used to segment the face skin area,and the morphological method is used to erode the eye and mouth areas between the skin,and the KCF algorithm is used to track the face,and the face area is divided into 2×2 size,then the r PPG waveform is extracted by the CHROM algorithm.Finally,the rPPG waveform is processed by the cross sliding window method to obtain the real-time changing heart rate.The experimental results show that the average accuracy between heart rate calculated by this method and heart rate measured by pulse oximeter is97.9217%,Bland-Altman consistency coefficient is 0.962302,which fully meets the clinical requirements.

关 键 词：远程光电容积描记法(rPPG) 人脸检测模型SSD300 OTSU算法 KCF算法 交叉滑动窗口模型(CSWM) 非接触心率测量 

分 类 号：TP261.1[自动化与计算机技术—检测技术与自动化装置]