TY - GEN
T1 - Analysis of factors on BVP signal extraction based on imaging principle
AU - Zhang, Xiaobiao
AU - Feng, Xiaoyi
AU - Xia, Zhaoqiang
N1 - Publisher Copyright:
© 2019 Association for Computing Machinery.
PY - 2019/5/29
Y1 - 2019/5/29
N2 - The extraction of human physiological based on face video has become a hot research direction, but few researchers pay attention to the process of signal extraction from an optical perspective. This paper establishes an optical model for human skin, and analyzes the principle of extracting human physiological signals from face video by imaging. Based on this model, the effects of melanin and hemoglobin on BVP (blood volume pulse) signal extraction were analyzed. In addition, this paper introduces the model of camera imaging, and discusses the cause of noise generation, analysis of the impact of two kinds of noise on video quality. Finally, this paper uses the MAHNOB database to carry out the heart rate extraction experiment from video, the experiment concluded that: (1) The face with lighter skin color is more conducive to human heart rate extraction; (2) Reducing the signal-noise ratio through compressing the video quality, the impact on the heart rate estimation error reaches minimum as the signal-noise ratio is 30dB; (3) By compressing the original video resolution, color information in the video is reduced, which has a impact on BVP signals extraction, the information recovery can be performed by the method of super-resolution reconstruction.
AB - The extraction of human physiological based on face video has become a hot research direction, but few researchers pay attention to the process of signal extraction from an optical perspective. This paper establishes an optical model for human skin, and analyzes the principle of extracting human physiological signals from face video by imaging. Based on this model, the effects of melanin and hemoglobin on BVP (blood volume pulse) signal extraction were analyzed. In addition, this paper introduces the model of camera imaging, and discusses the cause of noise generation, analysis of the impact of two kinds of noise on video quality. Finally, this paper uses the MAHNOB database to carry out the heart rate extraction experiment from video, the experiment concluded that: (1) The face with lighter skin color is more conducive to human heart rate extraction; (2) Reducing the signal-noise ratio through compressing the video quality, the impact on the heart rate estimation error reaches minimum as the signal-noise ratio is 30dB; (3) By compressing the original video resolution, color information in the video is reduced, which has a impact on BVP signals extraction, the information recovery can be performed by the method of super-resolution reconstruction.
KW - BVP signal
KW - Face video
KW - Skin color
UR - http://www.scopus.com/inward/record.url?scp=85072010746&partnerID=8YFLogxK
U2 - 10.1145/3345336.3345342
DO - 10.1145/3345336.3345342
M3 - 会议稿件
AN - SCOPUS:85072010746
T3 - ACM International Conference Proceeding Series
SP - 48
EP - 55
BT - Proceedings of 2019 3rd International Conference on Biometric Engineering and Applications, ICBEA 2019
PB - Association for Computing Machinery
T2 - 3rd International Conference on Biometric Engineering and Applications, ICBEA 2019
Y2 - 29 May 2019 through 31 May 2019
ER -