Facial Tracking in Thermal Images for Real-Time Noncontact Respiration Rate Monitoring

Abstract

A noncontact respiration rate monitor is developed. The method detected and tracked the face and then matched a pre-selected template in order to track a facial point of interest associated with respiration, located beneath the nostril. A region of interest (ROI) was then specified around the detected point of interest. Unwanted noise in the image was removed using image processing techniques. Feature extraction and signal processing techniques were applied to this ROI to compute the respiration feature and thereafter the respiration rate. It was shown that the method could detect and continuously track the subject's face by enclosing it in a rectangle. The method updated the location of the ROI and also determined respiration rate in a noncontact manner. The computation time for each frame was 40 ms, making it suitable for real-time respiration monitoring. Further work is in progress to enhance the algorithm to eliminate the need for pre-selection of a template.