A Study on Activity Recognition based on Temporal Change of the Temperature Distribution obtained from a Far-Infrared Sensor Array

Recently, the number of single-living elderly people is increasing along the aging of our society. Thus, there is a growing interest in systems that monitor them that can recognize daily and abnormal activities, while preserving their privacy. Here, daily activities include walking, sitting down, and standing up, and abnormal activities include falling. In this research, we propose an activity recognition method using the temperature distribution image obtained by a low-resolution far-infrared sensor array. A conventional method could not recognize the difference between activities where motion durations are similar, such as sitting down and standing up. The proposed method recognizes such activities using the features on trajectory, shape, and temperature of the human body region segmented from the temperature distribution image by background subtraction.