I. Introduction
Location Based Services (LBS) are mobile applications which rely on a user's location to deliver context aware functionality. Industry forecasts for this area predict huge market growth and revenue. One of the key issues for LBS is the positioning. Estimating the position of people and tracking is a fundamental challenge that has been investigated by many researchers for many years. One of the most popular positioning systems is Global Positioning System (GPS) [1], which has been widely used for positioning purpose in outdoor environment. However, GPS performs poorly in indoors and is not suitable for indoor positioning because of the following problems: (a) weak signal reception, (b) missing line-of-sight between the user and the satellites, (c) radio wave multi-path effect, and (d) scattering and attenuation in indoor environments. Thus, different wireless indoor positioning systems have been developed for indoor applications. In [2], different technologies such as infrared, ultrasound, RFID and Ultra wide band techniques are used. In addition, accuracy, coverage, measurement methods, and typical applications of those techniques are also described in [2]. Generally, there are three important technologies for indoor location positioning: (i) Infrared (IR) based systems that require line-of-sight path between users and transmitters and have limited applications, (ii) Ultrasonic based systems that localize a user based on the propagation time of ultrasonic wave. Such systems can achieve an accuracy within a meter. However, it needs large-scale basic infrastructures resulting in extra costs, and (iii) Radio frequency (RF) based systems. This is the most cost effective choice for indoor positioning because wireless local area network (WLAN) is available in most buildings and it does not require extra hardware; besides it has relatively large coverage. Among these positioning systems, the last one has been selected to explore its implementation strategies with machine learning.