Research on Crowdsourcing network indoor localization based on Co-Forest and Bayesian Compressed Sensing

Abstract

Indoor Localization Technology (ILT) based on Wi-Fi network has been rapidly developed with high localization accuracy and low hardware requirements. Collecting the Received Signal Strength (RSS) samples to construct the fingerprint database, however, is time consuming and labor intensive, hindering the application of the technology. An Indoor Localization Method is proposed based on Co-Forest and Bayesian Compressed Sensing (ILM-CFBCS), utilizing the crowdsourcing network technology to collect RSS data and the min-max method to preprocess the data so as to establish the indoor fingerprint database. The user's position is determined according to the decision result of the random forest classifier trained by the Co-Forest algorithm combining with majority principle. Finally, a constructing method of offline fingerprint database is put forward by combining the similarity between Bayesian compressed sensing theory and reference point fingerprint to realize the fingerprint database update. The experimental results show that the proposed method can achieve good localization performance by using a small amount of data with labeled positions.