Medicine Intake Detection Using a Wearable Wrist Device Accelerometer

Abstract. Medicine intake detection and feature selection algorithm are proposed in this paper. A precise medicine intake detection method is crucial in medicine intake monitoring system; it is an important factor in order to prevent the harm in the case of medicine non adherence for elderly patients or any people with chronic diseases and need to take medicine regularly. Wrist movement is the main function of intake gesture detection using a wearable 3-axis accelerometer. Activities of interest for medicine taking monitoring are drinking gesture, picking and taking medicine by hand and palm. This paper applied the eigenvalues and covariance for change detection and features selection. Eigenvalues are considered to solve the improper data truncation issue caused by a fixed window size, because respective time during those activities is very varying. Its results are satisfactory. The gestures were classified by feature characteristics with an accuracy as high as 96.7 % in the case of medicine taking by palm