A Mathematical Model for Fall Detection Predication in Elderly People

The falling risk of elderly people has become a significant issue. The use of a single sensor to detect the falling was found ineffective. Hence, methods such as video detection and use of more sensors were investigated, and falling prediction based on human body kinematics (HBK) and motion was studied. The model consisted of two algorithms: a prediction algorithm to predict the occurrence of a fall from daily activity living (DAL) and a decision-making algorithm to classify the DAL (fall or no fall). The model was analyzed using three inertial measurement unit (IMU) sensors with three degrees of freedom (DOF) that were assumed to be set on the thoracic, hip, and knee joints. The model used quaternions to represent the orientation of the three joints. To determine the occurrence of a fall, the joint angles for the thoracic, hip, and knee were calculated, and the world frame was used as a reference and a T-pose skeleton for coordinate calculation. The proposed model was evaluated using a ready-made dataset called IMU dataset; which contains real-time human motion obtained from IMU sensors. The evaluation was done using MATLAB simulation. The outcomes of the evaluation show that the proposed model is efficient and promising.