An Effective Algorithm for Detecting and Eliminating Wi-Fi Fingerprint Outliers

摘要

With the increasing popularity of intelligent devices, the application of indoor localization in all aspects is becoming more extensive. Indoor localization algorithm based on crowdsourcing can not only acquire exact locations but also eliminate the complex spot exploration process. However, the existing localization algorithms based on fingerprint are not resistant to the outliers, such as the indoor layout's changes, access point(AP) attacks. We proposes a robust algorithm named "Outlier Detection Based on Similarity"(ODBS) to detect outliers and eliminate the locating error due to them. The ODBS leverages a variety of sensors (e.g., accelerometer, compass, and gyroscope) built into modern smart phones to collect fingerprints on the paths while continuously scanning wireless signals. This paper realizes ODBS through the following processes: (a) Classify Wi-Fi fingerprints; (b) Fuse fingerprints; and (c) Find the possible outliers. The experimental result shows that ODBS is an algorithm with high robustness and accuracy to outlier detection.