TOWARDS PEOPLE INDOOR LOCALIZATION COMBINING WIFI AND HUMAN MOTION RECOGNITION Jos´ e M. Alonso 1 Alberto Alvarez 1 Gracian Trivino 1 N. Hern´ andez 2 F. Herranz 2 Manuel Oca˜ na 2 1 European Centre for Soft Computing (ECSC), Mieres (Asturias) {jose.alonso,alberto.alvarez,gracian.trivino}@softcomputing.es 2 Department of Electronics, University of Alcal´a (UAH), Madrid {noelia.hernandez,fernando.herranz,mocana}@depeca.uah.es Abstract This work presents a general framework for people indoor localization. Firstly, a WiFi localization system implemented as a fuzzy rule-based classifier (FRBC) is used to deal with the intrinsic uncertainty of such envi- ronments. It consists of a set of linguistic variables and rules automatically generated from experimental data. As a result, it yields an approximate position at the level of dis- crete zones (room, corridor, toilet, etc). Sec- ondly, a Fuzzy Finite State Machine (FFSM) mainly based on expert knowledge is used for human motion (activity, body posture and step length) recognition. The goal is find- ing out whether people is (or not) moving, in which direction, at which pace, etc. Finally, another FFSM combines both WiFi localiza- tion and human motion recognition with the aim of obtaining a robust, reliable, and eas- ily understandable human-oriented localiza- tion system. Keywords: WiFi localization, Human mo- tion recognition, Fuzzy rule-based classifier, Fuzzy finite state machine. 1 INTRODUCTION People localization systems provide interesting appli- cations in many areas [12, 14, 16]: personal naviga- tion assistance, medical assistance, finding and rescu- ing emergency first responders, personal security, etc. We are mainly interested in security applications (for instance sending warnings when someone gets into a dangerous area in order to reduce the occupational health and safety risk) and/or people assistance (for instance helping elderly or handicapped people). There are three main types of localization systems: (1) satellite based systems, (2) local network based systems and (3) sensor based systems. First, the satellite based systems, e.g., Global Posi- tioning System (GPS) [9], are widely used in outdoor applications with great successful. However, they do not provide precise indoor localization (nor even in cities with high buildings), making this problem an open challenge. Second, local network based systems use the network infrastructure to estimate user’s location. There are systems based on pre-existing networks like ZigBee networks designed for home control applications [7]. However, the most used systems are based on WiFi networks. The main advantage opposite to satellite based systems is that they are able to provide indoor absolute localization. In contrast, the principal draw- back is the need of a complete network infrastructure in the whole building where we want to localize a per- son. Luckily, this technology is quickly growing of cov- erage. Currently, there are WiFi Access Points (APs) in most public buildings like hospitals, libraries, uni- versities, museums, etc. In addition, measuring the WiFi signal level is free even for private WiFi net- works. As a result, WiFi technology is a good choice for indoor global localization systems yielding a good accuracy-cost trade-off [2]. Third, sensor based systems provide absolute infor- mation (e.g., magnetic compass, ultrasonic or infrared sensors) or relative information (e.g., inertial measure- ment units or pressure sensors). One low-cost inertial sensor is the accelerometer, based on the Micro Elec- tro Mechanical Systems (MEMS) technology that has allowed its integration in small and low energy con- sumption devices. Accelerometers can be used as step length estimators; furthermore they let us to obtain some information about body posture [21]. Human activity can also be analyzed by means of combining one accelerometer with a skin conductivity meter [22]. ESTYLF 2010, Huelva, 3 a 5 de febrero de 2010 XV Congreso Español Sobre Tecnologías y Lógica Fuzzy 7