Dataset Descriptions

This dataset contains facial videos of 9 subjects and corresponding reference contact PPG sensor data, which were used for the evaluation of inter-beat interval (IBI) estimation in our EMBC2019 paper [Project]. This dataset is available only for academic use and research purpose.

• Experimental Setpups
  Camera:
    Used camera: Prototype RGB-NIR camera [Detail]
    Framerate: 300fps (original) and 30fps (every 10 original frames are averaged)
    Resolution: VGA (640*480)
  Illumination:
    2 KINOFLO lights (FIX-4804/FIX-2404 Series) and 2 NIR lights (SunMechatronics L-light)
  Refarence BVP:
    Contact PPG sensor (Procomp Infinity T7500M, Thought Technology Ltd., Canada)
  Subjects:
    9 subjects (8 male and 1 female) with age 20s - 60s


• Experimental Procedure
  Each video contains three sessions: relax (60[s]), exercise (60[s]), and relax (60[s]) sessions. In the exercise session, the subjects were asked to perform handgrip exercise.
  
  

  

  
  

  

  
  
• Dataset Contents
  Videos:
    Nine 20-second videos for each subject (We divided each 180-second video into 9 fragments)
  Original images:
    If you also need original images of each frame (19GB PNG files for each subject, in total 173 GB for all 9 subjects), please e-mail to vital-sensing[at]ok.sc.e.titech.ac.jp.
  Facial landmark files:
    Facial landmarks detected with dlib. (GitHub: "find_face_landmarks")
  Contact PPG sensor data:
    PPG data recorded by a finger-attached contact PPG sensor. The sampling frequency of the contact PPG sensor is 2048[Hz]. The detected peak locations by our algorithm is also provided.
      -dataA: PPG amplitude
      -locsA*: detected peak location
      -pksA* : detected peak height
        (*Detected by our algorithm.)
  Note on synchronization:
    Video and contact PPG (cPPG) sensor are not strictly synchronized, but we experimentally synchronized them using a flash light as follows.
  
  timeStamp_Video = timeStamp_cPPG - 0.32 [sec.]
  
  
• Download Procedure
  In order to use our dataset, you have to agree with End User License Agreement (EULA). Please carefully read and then send the signed EULA to vital-sensing[at]ok.sc.e.titech.ac.jp. After that, you will receive a download link by e-mail.
  


• Citation
  When you use the dataset for future publications, please cite the following paper.
  
  "Inter-Beat Interval Estimation from Facial Video Based on Reliability of BVP Signals,” Yuichiro Maki, Yusuke Monno, Kazunori Yoshizaki, Masayuki Tanaka and Masatoshi Okutomi,
  International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC2019), pp.XXXX-XXXX, July 2019 (to appear). [PDF]