[1]李 伟,杨 庚.基于 LBF 的人脸视频心率测量[J].计算机技术与发展,2020,30(12):123-129.[doi:10. 3969 / j. issn. 1673-629X. 2020. 12. 022]
　LI Wei,YANG Geng.Facial Video Heart Rate Measurement Based on Local Binary Features[J].,2020,30(12):123-129.[doi:10. 3969 / j. issn. 1673-629X. 2020. 12. 022]
点击复制

基于 LBF 的人脸视频心率测量()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
30
期数:
2020年12期
页码:
123-129
栏目:
应用开发研究
出版日期:
2020-12-10

文章信息/Info

Title:
Facial Video Heart Rate Measurement Based on Local Binary Features
文章编号:
1673-629X(2020)12-0123-07
作者:
李 伟1; 2; 杨 庚1; 2
1. 南京邮电大学 计算机学院、网络空间安全学院,江苏 南京 210023; 2. 江苏省大数据安全与智能处理重点实验室,江苏 南京 210023
Author(s):
LI Wei1; 2; YANG Geng1; 2
1. School of Computer Science and Network Space Security,Nanjing University of Posts and Telecommunications,Nanjing 210023,China; 2. Jiangsu Key Laboratory of Big Data Security and Intelligent Processing,Nanjing 210023,China
关键词:
心率测量; 光电容积描记技术; 人脸对齐; 局部二值特征; 色度模型
Keywords:
heart rate measurement; PhotoPlethysmoGraphy (PPG); face alignment; local binary features; chrominance model
分类号:
TP391
DOI:
10. 3969 / j. issn. 1673-629X. 2020. 12. 022
摘要:
心率是衡量人体健康和心理状况的重要指标,针对目前基于人脸视频的非接触式心率信号的提取易受人脸倾斜的影响,提出了一种基于局部二值特征的人脸视频心率测量方法。 首先对使用普通摄像头采集的人脸视频进行人脸检测,运用局部二值特征模型实现人脸对齐,在面部区域精确选取感兴趣区域(ROI);其次在 HSL 颜色空间下对 ROI 进行皮肤分割,利用色度模型进一步去除脸部运动的干扰,并通过巴特沃斯带通滤波器减少噪声影响,提取脉搏信号;最后对提取的脉搏信号利用离散傅里叶变换进行时频分析得到心率值。 实验结果表明,基于局部二值特征的人脸视频心率测量方法在人脸倾斜的情况下仍有较好的准确性和稳定性,平均绝对误差为 2.205 0,均方根误差为 2.639 5,与指尖脉搏血氧仪测量法具有很好的一致性,与其他方法比较性能有所提升,能够满足日常生活和远程医疗中心率测量的需要。
Abstract:
Heart rate is an important indicator for measuring human health and psychological conditions. Concerning the problem that the extraction of non- contact heart rate signals based on facial video is susceptible to face tilt, a method of facial video heart rate measurement based on local binary features is proposed. Firstly,face detection on the video collected by the ordinary camera is performed and face alignment is implemented by the local binary features model to select the region of interest (ROI) accurately in face area.Secondly,skin segmentation is performed on the ROI in the HSL color space. In order to extract pulse signals,the chromaticity model is used to further remove the interference of facial motion, and the Butterworth band-pass filter is used to reduce noise impact. Finally,the measurement of heart rate is implemented by time-frequency analysis through discrete Fourier transform of the extracted pulse signal. The experiment shows that mean absolute error (MAE) is 2.2050 and root mean square error (RMSE) is 2.639 5 when the face is tilted. Meanwhile,the facial video heart rate measurement method based on local binary features is largely consistent with the fingertip pulse oximeter measurement method and improved in performance compared to other methods,which can meet the needs of daily life and telemedicine center rate measurement.
更新日期/Last Update: 2020-12-10