Design And Implementation Of Abnormal Situation Awareness System For Elderly People

Internet of Things (IoT) is becoming one of the hottest research topics. Thereare various applications by using IoT technologies, e.g. smart home, smart grid, andsmart city etc.Nowadays, many closely related people are separated by distance because ofwork, study, and some elderly people are living alone with not much connection andhealth care supports from his/her remote family. Thus, these remote families areeager to know each other’s daily life. In our view, some abnormal daily lifeinformation can be sent to his/her remote family, so that the remote family can paymore attention to the situations, and finally their interaction can be enhanced. Butsharing daily life information will cause some privacy vs. awareness conflicts, e.g.the elder user doesn’t want his/remote family be aware of too much information, butthe remote family really want to know; or the elder user wants his/her family to paymore attention to his/her situations, but his/her remote family feel annoyed with toomuch information.To solve above problems, in this thesis, we propose a support method ofmutual situation-aware for elders and remote family members. There are three mainparts in our support method:(1) the daily life situation of a user living in smarthomes can be detected;(2) the situation can be classified as normal or abnormal;(3)the remote family can be aware of the abnormal situations by receiving theinformation from smart phone after a mediation algorithm for resolving privacy vs.awareness conflicts. For the1stpart, we detect user’s zone-based situations by usingU-tiles, a sensors network for location detection, and some other devices likeUbi-command Human Interface (UHI), a device which can detect user’s behaviorslike eating, running and so on, through detecting user’s finger gestures. For the2ndpart, for every situation, we choose some meaningful and significant features to useSupport Vector Data Description (SVDD), a one-class machine learning method bysearching a closure boundary on the dataset, to learn a user’s situation and classifythe situation as normal an abnormal. And for the3rdpart, we try to solve user’sprivacy vs. awareness conflicts problem by a mediation algorithm based on theclassification results of SVDD method. The elder user and his/her remote familycan select different acceptable boundary values on normal and abnormal, and according to their choices, we mediate the SVDD boundary to solve the conflicts tomake both of them be satisfied. Our method is implemented in Java, MATLAB andAndroid, and an experiment is performed. And we make some questionnaires toevaluate our method.