Wandering Detection and Activity Recognition for Dementia Patients Using Wireless Sensor Networks

Abstract

A central goal of information technology is to improve human life. In terms of useful technology in the area of sensor networks, activity recognition (AR) has become a key feature. Using AR technology, it is now possible to understand human behavior, including what, how and when people perform an activity. In recent years, there has been an increase in accident reports involving aged dementia patients, resulting in higher social costs to treat and care for dementia patients. AR technology can be utilized to take monitor the activity of these patients. In this paper, we present an efficient method that converts raw sensor data to readable patterns in order to classify an individual's current activities and then compare these patterns with previously stored patterns to detect any abnormal patterns, such as wandering, which is one of the early symptoms of dementia. We used this method to digitize human activities and applied the Levy-walk model to detect wandering patterns. We developed an inference model for an early dementia symptom based on digitized human activity patterns. In this article, we also illustrate the implementation of a sensor system configured with tri-axis acceleration and an ultrasonic sensor as well as a wandering estimation algorithm in order to overcome limitations of existing models to detect/infer dementia.