A Vision Based approach for Fall Detection System for Elderly People

Abstract

There is a surge in the aging population, it is important for doctors and guardians to be conscious of issues related to health, which could impact majorly to the health of the elderly. Falling is one of the significant concerns that lead to fatal health problems in the case of the older population. The proposed system is especially for elderly people who are living alone in their homes or for old age homes. A vision-based fall detection approach has been used below in a home environment. A UR dataset of kinetic sensor’s output videos is used which classify into two categories i.e. fall and activities of daily living which are considered for experimentation. The MOG2 algorithm is used for background subtraction to focus on the human object and ignore the rest of the surrounding environment. The Shi Tomasi algorithm is used for finding interest points that are tracked using the optical flow method: Lucas-Kanade algorithm. Using interest points we compute its maximum displacement along with direction and speed of motion. For fall activity the last positions of interest points were observed. So the system detects a person in the environment, keeps tracking the person, and calculates the optical flow in case of a fall. If the fall is above the threshold point, an alert is sent to the concerned people and medical authorities.