Accurate Hemodynamic Sensing using Video Plethysmography with High Quality Cameras

Facial video plethysmography provides non-contact monitoring of heart activity based on blood volume pulsations in facial tissue. Traditionally, the video is captured using a simple camera followed by sophisticated signal processing methods. This approach provides accurate average heart rate but falls short in providing accurate beat to beat detection and irregular heartbeats. In this work, we explore the possibility of accurately sensing beat to beat plethysmography signals using a high-quality image sensor camera coupled with conversion to color spaces where the cardiac pulse is most pronounced. We evaluate our approach in a clinical study by gathering data from 60 subjects in a hospital setting using an array of sensors. We use the data to assess accuracy of inter beat interval detection and heart rate variability estimation compared to those obtained from an earlobe photoplethysmography sensor and from an electrocardiogram sensor. In addition, we investigate the possibility of identifying irregular cardiac activity such as individual ectopic beats. For additional comparison, we also apply blind source separation to the data. Our results demonstrate that coupling a high-quality camera with the appropriate color conversion senses a beat to beat signal equivalent to that obtained from a skin photoplethysmography sensor. This signal provides accurate estimation of inter-beat intervals, heart rate variability parameters and identification of irregular cardiac activity. These results demonstrate how combining a high quality camera with the appropriate color conversion provides a complete hemodynamic sensor.