Device Free Indoor Localization Using Discriminant Features of CSI A Canonical Correlation Paradigm

Abstract

With the growth of context-aware computing, Indoor localization based on Wi-Fi signals is gaining popularity due to the widespread deployment of Wi-Fi infrastructure. In this paper, we propose DisLoc, the first discriminant feature based device free indoor localization using canonical correlation analysis. The system exploits the Channel State Information (CSI), which is richer in information related to position than traditional Received Signal Strength (RSS). The DisLoc models localization as a pattern matching problem and addresses the problem of comparing sets of CSI feature matrix for indoor 10-calization, where the sets represent variations in features in terms of CSI amplitude due to the presence of a subject at a particular location. A discriminative feature extraction is proposed for pattern matching using Canonical Correlation Analysis (CCA). Specifically, the system utilizes Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) of CSIs as two features into the correlation analysis and performs an effective feature fusion by maximizing the pairwise correlations across the two feature sets. Finally a similarity measure is performed to find the best match for localization. Experimental results show that DisLoc can estimate location in a device free setting with high accuracy which outperforms other state-of-the-art approaches.