Fingerprint-Based Device-Free Localization in Changing Environments Using Enhanced Channel Selection and Logistic Regression

Abstract

Device-free localization (DFL), which can detect and locate a person by measuring the changes in received signals, is one of the primary techniques in wireless sensor networks. Recently, research on fingerprint-based localization in changing environments has been receiving increasing attention. However, when the environment changes due to furniture or other objects are moved, there is still much room for localization accuracy improvement in fingerprint-based DFL. In this paper, we propose a novel DFL algorithm for changing environments: this algorithm features an enhanced channel-selection method and adopts the logistic regression classifier to improve the localization accuracy. The proposed frequency channel-selection method selects two correlated channels with higher Pearson correlation coefficient both in the training and testing procedures, which would be more robust to the environmental change. Meanwhile, the logistic regression classifier could counteract the negative influence on the localization accuracy, without the need for rebuilding the database in fingerprint-based DFL. Experimental results demonstrate that the logistic regression classifier has the lowest error rate among three related methods (k-nearest neighbours classifier, linear discriminant analysis classifier, and random forests classifier). In addition, the localization accuracy has been further improved by the proposed DFL algorithm than by the other state-of-the-art fingerprint-based methods.