Study of respiration measurement method by optical flow considering body motion

Abstract

Measuring daily vital signs unconsciously is useful for early detection and prevention of diseases. We have proposed to measure the pulse rate and respiration rate during sleep noncontactly by using the near infrared image. Since body movements occur during sleep, a method of determining an appropriate calculation position of optical flow was studied. Optical flow and reliability at each position of the entire image were calculated using the Lucas-Kanade method. Next, the x component and the y component of the optical flow at each position were calculated, and a low pass filter was applied to this to obtain a respiration waveform. Near-infrared images during sleep while wearing a respiration sensor was recorded. As a result of investigating the data between 150s before and after turning over, respiration could be detected with accuracy of about 90% by using respiration waveform in shoulder region and high reliability region.