Writing‐State Classification by Ambient Sedentary Behavior Sensing in Desk Work

Abstract

SUMMARYThis paper presents a technique for writing-state classification by sedentary behavior sensing with a sensorequipped chair, which wirelessly monitors the user's states such as behaviors or physiological and psychological conditions, for estimation of the user's subjective difficulty with a studying problem on a desktop. Avoiding user discomfort, four load cells are mounted behind the seat plate of a regular office chair. The system simply measures the weight and the center of pressure of the seat. These data are divided into time segments, which are labeled into four sedentary-body sway primitives by the decision tree algorithm, and then the behaviors are derived from the number and pitch of those primitives in a certain interval. In systemevaluation experiments, it achieved more than 80% labeling accuracy for untrained users. We conducted a user experiment showing its potential for desk work applications. The results of the experiment showed that our proposed system could classify user states into the writing state and other states such as reading a document and watching a movie at a classification rate of 87%. C⃝ 2014 Wiley Periodicals, Inc. Electron Comm Jpn, 98(1): 15-22, 2015; Published online in Wiley Online Library (wileyonlinelibrary.com).