An automatic approach to fingerprint construction of indoor localization by crowd paths

Abstract

In typical indoor position system employing location fingerprints model, received signal strength indications (RSSI) from a set of Wi-Fi access points are used as an unique fingerprint to identify a specific position. This type position systems need abundant Wi-Fi fingerprints, generally implemented by trained experts, which extends labor costs and restricts heir promotion. In his paper, a novel approach to construct intelligently Wi-Fi fingerprint database based on crowd paths triggered by lots of ordinary users holding onto smartphone is proposed. As existing drift errors in inertial measurement unit (IMU) and Wi-Fi module and the knowledge that crowd paths involve massive similar or crossing positions, we defined a concept: thin landmarks, and employ a fuzzy voter scheme to locating each thin landmark. Then rectifying the position of each sample point in every thin landmark's candidate set with corresponding thin landmark's coordinate and simultaneously utilizing particle filter (PF) algorithm to smooth each sample point on each crowd path. We implemented the approach on off-he-shelf smartphones and evaluate he performance in our campus. Experimental result indicates that the approach can availably construct Wi-Fi fingerprint database in tolerable errors.