Indoor positioning system using software defined radio (SDR) and support vector machine (SVM) through use of orthogonal frequency division multiplexing (OFDM) signals
Indoor Wi-Fi positioning systems (WPS) are useful for location determination and rely upon the use of existing Wi-Fi hotspot. These systems are advantageous to Global Positioning System (GPS), another popular location positioning system, when GPS is inadequate due to various causes including multi-path and signal blockage. WPS systems measure intensity of the received signal and also applying "fingerprinting" techniques to determine location; however, these systems are only as accurate as the number of locations and number of concurrently received signals from access points that have been entered into a database. In addition, fluctuations in signals due to changes in environment can also have a detrimental effect on accuracy. Understanding the causes that lead to poor indoor positioning system performance as well as techniques that can be used to improve it is very important. Software defined radio (SDR) is an emerging, state-of-the-art technology which features modulation/de-modulation and other techniques in digital signal processing (DSP) in software as opposed to hardware. SDR technology is also very useful in areas where evaluation and analysis of radio frequency (RF) signals is needed. Due to its extreme flexibility, SDR can be modified quickly. SDR lends itself well to the use and evaluation of other deterministic classification techniques which, when applied to RF signals, can be useful for location determination. One such classification technique is Support Vector Machine (SVM) technology. This paper exploits signal propagation multipath as fingerprints in environments with only few available access points which typically result in poor WPS performance. Multipath fingerprints are then used in implementation of an indoor positioning system using a SDR architecture and SVM classification. The SDR architecture is comprised of a popular SDR platform - GNU Software Radio with a new Universal Software Radio Peripherals (USRP) device - USRP Network Series N210. Classification is performed using a popular classification technique - Library SVM (LibSVM). The SDR architecture and classification leverage Orthogonal Frequency Division Multiplexed (OFDM) radio frequency (RF) signals transmitted from another GNU Radio/USRP SDR platform. The study demonstrates the feasibility of using multipath fingerprints for WPS-like positioning that justifies further research on providing better ways to understand the causes that lead to poor indoor positioning system performance as well as techniques that can be used to improve it.