HyLoc: An Indoor Single-Site Hybrid Localization Scheme Based on LOS/NLOS Identification

Abstract

The complex indoor structure not only introduces line-of-sight (LOS) paths but also non-line-of-sight (NLOS) paths, which poses a huge challenge to localization. However, most of the existing indoor localization schemes only utilize a single positioning algorithm for LOS or NLOS environments, resulting in poor positioning robustness. To solve this problem, we propose an indoor single-site hybrid localization scheme called HyLoc in this paper. HyLoc combines multiple positioning algorithms and gives full play to the advantage of each algorithm in either LOS or NLOS environment. In this scheme, a threshold judger (TJ) is firstly designed to identify whether there is a LOS path depending on the time-domain statistical features extracted from channel state information (CSI). According to the identification results of TJ, HyLoc adaptively selects the optimal positioning algorithm. In the LOS environment, an improved multiple signal classification algorithm (MUSIC) based on forward smoothing technology is applied to obtain the estimated positioning results. In the NLOS environment, a multipath subspace projection and extreme learning machine (ELM)-based fingerprint localization algorithm is proposed for positioning analysis. Finally the experimental results verify that the proposed HyLoc can realize single-site localization and it has higher positioning accuracy than traditional ones in the mixed LOS and NLOS environment.

Keywords

• Indoor single-site hybrid localization
• LOS/NLOS identification
• multipath signal subspace
• extreme learning machine