Extreme Learning Machine Based Location-Aware Activity Recognition

Abstract

According to the recent government reports, China has gradually entered an aging society. Pension problem is a vital problem to face. Therefore, it will be very useful to monitor the health status of elderly people who live alone at home. To evaluate the abilities of elderly people in daily life, the activities of daily living (ADL) is used. In this work, we propose a novel machine learning approach for ADL recognitions by considering the location context information of the elder. With the popularity of smart phones, motion recognition can be done by the embedded sensors such as acceleration sensors and. However different ADL models possibly have the same movement to a certain degree, which will affect the classification performance. We append the location information as an additional feature to detect ADL. Furthermore, we propose a hierarchical Extreme Learning Machine (ELM) to classify the ADL. With the experiment and test, the algorithm described in this paper can achieve obvious performance in ADL recognition.