Field Experiments of a Sensor Embedded Medication Support System

In this study, the authors report results of month-long experiments conducted in two elderly recipients' houses with a new reminder system to keep recipient medication condition up to date. This system consisted of three main components: a new intelligent medicine storage case, wireless sensing devices, and a database server. The medicine case recognized the presence of medicine in its storage compartments by image processing and estimated the recipients' present living state (i.e., current activity) on the basis of a fuzzy inference algorithm. The sensing devices were distributed throughout the houses to collect big data for living state estimation. The server accumulated results of calculation for remote monitoring of medication compliance and collection of historical data. From the experiments, the authors concluded that the proposed system was more effective in inappropriate medication intake time prevention than conventional systems based on warning accuracy. The rate of medicine recognition was 100%, and the overall F-measure of living state estimation was approximately 70%. As a result, the proposed system could detect a mistake when an elderly recipient forgot to take medicine at the prescribed time.