Remote Detection of Photoplethysmographic Systolic and Diastolic Peaks Using a Digital Camera

• Published in: IEEE transactions on biomedical engineering Vol. 61; no. 12; pp. 2948 - 2954
• Main Authors: McDuff, Daniel, Gontarek, Sarah, Picard, Rosalind W.
• Format: Journal Article
• Language: English
• Published: United States IEEE 01.12.2014; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Adolescent; Adult; Algorithms; Biomedical measurement; Blood Pressure - physiology; Blood Pressure Determination - methods; Blood Volume - physiology; Blood Volume Determination - methods; Diagnosis, Computer-Assisted - methods; Diastole; Digital cameras; Face - anatomy & histology; Face - blood supply; Face - physiology; Female; Heart rate variability; Humans; Male; Morphology; Photography - methods; Photoplethysmography; Photoplethysmography - methods; Reproducibility of Results; Sensitivity and Specificity; Systole; Video Recording - methods; Young Adult; Blood volume pulse; morphology; remote; heart rate variability
• ISSN: 0018-9294; 1558-2531
• DOI: 10.1109/TBME.2014.2340991

We present a new method for measuring photoplethysmogram signals remotely using ambient light and a digital camera that allows for accurate recovery of the waveform morphology (from a distance of 3 m). In particular, we show that the peak-to-peak time between the systolic peak and diastolic peak/inflection can be automatically recovered using the second-order derivative of the remotely measured waveform. We compare measurements from the face with those captured using a contact fingertip sensor and show high agreement in peak and interval timings. Furthermore, we show that results can be significantly improved using orange, green, and cyan color channels compared to the tradition red, green, and blue channel combination. The absolute error in interbeat intervals was 26 ms and the absolute error in mean systolic-diastolic peak-to-peak times was 12 ms. The mean systolic-diastolic peak-to-peak times measured using the contact sensor and the camera were highly correlated, <inline-formula><tex-math>\rho </tex-math></inline-formula> = 0.94 ( p <inline-formula><tex-math><</tex-math></inline-formula> 0.001). The results were obtained with a camera frame-rate of only 30 Hz. This technology has significant potential for advancing healthcare.