Improvement in Accuracy of Respiration Pattern Detection on Visual Sensing System

Non-restrictive monitoring of respiration of a patient is one of the most important tasks in the field of clinical care. We have already extracted frequency characteristics by interimage subtraction from motion pictures that show motion of respiration (Visual Sensing System). However, there is a limit to elucidating physiological changes using motion pictures, due to S/N ratio, frame rate of video camera, and drift caused by salt and pepper noises. Our new work contains three algorithms to extract frequency characteristics optimally despite these limitations. The first is adaptive intenpal setting, in which we most adaptively set the interval for inter-image subtraction to optimally detect respiration patterns. The second is adaptive threshold setting, in which a threshold is most adaptively set for counting cross points so that respiration patterns on the curves, large or small, are not ignored. The last is time averaging of image data, which reduces the effects of salt and pepper noises. As a result, frequency characteristics in the darkness are recognized despite salt and pepper noises. This study improves the quality of respiration monitoring system that is an indispensable apparatus to keep patients' life.