A depth image approach to classify daily activities of human life for fall detection based on height and velocity of the subject

Abstract

Human fall detection systems can be categorized according to the approaches used such as some sort of wearable devices, ambient based devices or non-invasive vision based devices using live cameras. Wearable and ambient based devices are very often rejected by users due to the high false alarm and difficulties in carrying them during their daily life activities. This work proposes a fall detection system using depth information from Microsoft Kinect sensor. Classification of human fall from other activities of daily life is accomplished using height and velocity of the subject extracted from the depth information. Results of the simulated activities showed that the proposed system is able to gain an accuracy of 93.75% with 100% sensitivity and a specificity of 92.5%.