Activity Recognition for Risk Management with Installed Sensor in Smart and Cell Phone

Abstract

AbstractSmart and cell phone with self-contained sensor such as accelerometer, gyroscopic and digital magnetic compass sensor have been popular. Combining certain algorithm and those sensors, it can estimate user’s activity, situation and even user’s absolute position. However, estimation of user’s activity, situation and user’s absolute position become difficult when once sensors posture and position are changing from original position in user’s motion. Also, according to stored, worn and handheld position and posture of those cell and smart phone are often changed. Therefore, we exclude estimation of user’s position and we focus to only estimation of user’s activity and situation for risk management. Basically, we design special classifier for detecting user’s unusual behavior and apply other user’s position data from internet to the results detected by the classifier which are combined wavelet transform and SVM. We assume that user’s unusual activity and situation can be detected by smart and cell phone with high accuracy.KeywordsActivity recognitionWearable computerSVMnearly fall incidentcell phone and smart phone