Context Recognition Based On Mobile Devices' Data

In the past 10 years,the development of the mobile devices is dramatically rapid,especially the development of smart phone in the recent years.The mobile devices are designed to be more and more compact and portable,and in the meanwhile,the functions provided by the devices is increasing,a lot of APP or software are designed and developed,which makes the mobile devices become popular and increase uses' engagement,even make the mobile devices to be an inseparable part of people's life.At the same time,more and more sensors are integrated on the mobile devices,which makes it possible for us to collect the user's different types of contextual and environmental data.These data sets provide rich information for us to recognize the user's current context.The increasing number of APP and software makes user and devices become closer and closer,people rely on these comprehensive and convenient services much more than ever.Currently,most of the APPs and soft wares provide services regardless of the context of the user or the state of the user.However,a good software,especially the one has function related to the users' environment and context,should understand the information of users' location and current states,via data mining and analysis of the user's relative environmental and contextual data.It is an important part of personalized service,for instance,providing different services for users in different contexts,which brings pleasant user experience.Also,the data mining of mobile device data can help capturing the crowd motional information especially in these big scaled events,and understanding the individuals and crowds motive pattern.The data of mobile devices can be used to characterize and model users' activity and predict users' motion.It plays an important role in subjects such as Sociology,economics,urban planning.Therefore,our work makes use of users' contextual and environmental data from the sensors integrated in the mobile devices,and the data mining and analysis algorithms to complete the task of recognizing users' context.The main challenges of this work are as follows:1.Mostly,traditional methods of recognizing users' location are based on the sensor data from the mobile devices directly.The way to characterize the real world by sensor has a lot of drawbacks.For example,sometimes the characterization is not accurate enough,but sometimes it is too much fine-grained,and usually regardless of users' need.2.The definition of the concept “context” varies in different works.Some works treat the “context” as the location of the users,but others may think it to be the users' current motion or activities.3.In the related work about “context recognition”,researchers have provided several algorithms and proved the efficiency of the algorithms.But there is still lack of system framework to solve such problems.In summary,the main work of this paper is as follows: 1.This paper provides unified definitions of some relative concepts,such as personal property and context.2.We design a scalable three-layer model as the system framework.Sensors layer in the last,property layer in the middle,and context layer on the top.3.We provide the data mining algorithms of mapping from lower layer to higher one,and in this way to recognize the context.