This paper proposes a method to estimate the state of user to provide proactive hospitality from features of their gait pattern acquired with an Radio Frequency Identifier (RFID) system. This method uses RFID readers on each shoe, as well as RFID tags installed on the floor. The ID of each tag is organized as a map, to show the precise position of the user. The reader and tags communicate while the user is walking. We classify tag IDs detected by readers into each step with Ward Method. We calculate stride, walking speed, and so on, as feature components of a gait vector in each step. We recognize the state of the user from these components with the Random Forest. In an experiment, we have imposed subjects on walking under several kinds of conditions. We have evaluated the classification result through F-measure calculated from 10-fold cross-validation. It implies we can classify each state of users. We discuss why we can classify each state of users from gait vector components with the variable importance and their correlation. In addition, we have verified whether we can detect discomfort caused by the way to carry luggage. Finally, we discuss the feasibility of our proposed method.
展开▼
