Research On Mobile Device Based Indoor Localization Method

With the rapid development of wireless communication technology and pervasivecomputing technology, location based service (LBS) has gradually become one of hot topics,and is widely used in various fields of health care, underground rescue, intelligent home,social activities, etc. The key point of LBS is accurately achieving the user’s locationinformation. However, extensive business systems can get the location based on GlobalPositioning System (GPS) and telecommunication cellular networks(GSM/CDMA/WCDMA), they exist many shortcomings such as failing to obtain signalsindoors, privacy protection, large power consuming, poor localization accuracy and so on,which constrains their wide application and extension.With the development of economic construction, the area of indoor places is growinglarger and larger, and such places like large-scale shopping mall, office site and entertainmentplace are gradually emerging. In the our daily life, people mostly act indoors, conveyinggeo-coordinates as location information can’t be used in the indoor environment, which leadsthat people have great demand for indoor semantic location services. At present, Wi-Fimodules and Bluetooth sensors which are pervasively deployed in mobile devices make thesemantic location recognition feasible. This paper proposes HMM based indoor localizationmethod, which is on the basis of existing indoor localization principle of Wi-Fi fingerprint. Asfor the semi-heterogeneous feature problem of high fluctuation indoor Wi-Fi data, this methodcan well handle this problem and adaptively meet the demand of indoor high dynamicenvironment. Meanwhile, this paper utilizes Bluetooth sensors which are also pervasivelyembedded on mobile devices, to collect dynamic context information, then train the combinedlocalization model, so as to enhance the Wi-Fi localization effective.In detail, this paper mainly studies on dynamic Wi-Fi wireless signal and dynamicBluetooth information, and achieves the following aspects of research results:Firstly, as for the semi-heterogeneous feature of Wi-Fi data, this paper proposes hiddenmarkov model based localization method called LocHMM. This approach transformsextracted Wi-Fi signals into the sequence in terms of Wi-Fi received signal strength, andemploys HMM to train the localization model, then the average accuracy can reach92.6%forthe room-level localization.Secondly, we collect the Bluetooth information from typically semantic locations, such as office, meeting room, subway, restaurant, park and shopping mall, and extract noveldistinguish features, then establish the localization model using Decision Tree algorithm, thelocalization accuracy reaches87.8%averagely for these six semantic locations.Thirdly, dynamic Bluetooth information has very high discrimination ability, which isapplied to modify the Wi-Fi localization result and enhance the room-level accuracy to reach94.9%, and also supply location services for the coverage with no Wi-Fi signals, under thecircumstance of not depending on any other equipment.Finally, we design and implement a location based mobile commodity informationinteraction prototype system, also construct the location sample base by Bluetooth and Wi-Fiand relevant commodity information, so as to provide the researchers’ reference studied insome related research interests.