Deriving activity from RFID detection records in an assisted living context
There has been a significant growth in the deployment of Radio Frequency Identification (RFID) in the supply chain in the recent years. Since RFID requires less cost and infrastructure than the sensor networks such as Ultrasonic or Wi-Fi, it has been applied in many business and research domains. The major applications of RFID technology include location position, object tracking and activity monitoring. Particularly in the individual activity monitoring circumstance, it is expected that by detecting the object with which the person interacted, the related personal movement is able to be recognized. However, the inherent unreliability of RFID data results in the uncertainty of the RFID detection of activity. This research estimates the accuracy of using RFID detection records to monitor personal activity in an assisted living context. The methodology for this research is a quantitative approach by using design science. Two real experiments are conducted for performing activity monitoring in a laboratory environment. Both experiment results show that false positive reads have a serious influence on the accuracy of detecting individual motion. In order to remove the noisy data from the original RFID data stream, the multi-level data pre-processing method is used and analysed in the research. The cleaned dataset shows the perfect accuracy of personal activity inference. Although this proposed method is efficient for filtering the noisy data and predicting the correct individual movement in this research, it only focuses on recognizing the regular personal activity in a clean indoor environment. Due to the complexity of human indoor behaviour, future exploration needs to be carried out in other different environmental backgrounds.