A novel method of Wi-Fi indoor localization based on channel state information

Traditional Wi-Fi indoor localization system based on big data technique suffers from great degradation due to the instability and low space distinguish ability of received signal strength (RSS). Replacing RSS with channel state information (CSI) is proven to be an efficient method. However, not all CSI raw data contribute equally to the localization performance. The computational cost of fingerprint database is unacceptable as well. In this paper, we propose a novel method of Wi-Fi indoor localization based on CSI. A fast orthogonal search (FOS) algorithm is utilized to calculate the weights of CSI raw data collected from the wireless network interface card (NIC), reducing the database at the same time. Different weights of the features are then used as the input of a back-propagation (BP) neural network, conducting weighted training. We implement the system and experimentally evaluate its performance in the typical laboratory scenario. The performance of the proposed system is compared with several existing systems. Result shows that the proposed system has a 13% improvement in accuracy and a 14% improvement in execute time. The average distance error is 1.5702m.