Novel Indoor Positioning Mechanism Via Spectral Compression

Abstract

Received signal strength (RSS) measurements are important in indoor location solutions based on WiFi, cellular networks or Bluetooth. RSS-based positioning involves two phases, namely, learning and estimation. The database sizes required both for the learning and for the estimation phases grow rapidly as the network coverage areas and the number of access points number increase. Achieving large-scale/global localization solutions would be possible if the database size bottlenecks were solved. We present here an innovative approach based on spectral compression, which allows a tremendous reduction in the database sizes in both learning and estimation phases. We introduce the new concept of compressed RSS images. We show how, through an astute 2-D frequency analysis, only a fraction of the transform-domain components need to be stored and transferred to/from the mobiles. Our idea is validated with WiFi real-life measurements from five multistory buildings. We show that our method is able to provide comparable results with the traditional fingerprinting approach, but with up to 80% reduction in the database sizes.