Wi-Fi Fingerprint Localization Based on Multi-output Least Square Support Vector Regression

Estimating the location of a movable object is highly necessary for providing context-aware services in an indoor environment. As Global Positioning System (GPS) is not appropriate for indoor positioning, Wireless Local Area Network (WLAN) seems to be the choice due to its ubiquitous nature. The localization task based on wireless signals involves several challenges. This paper proposes a cost-effective Wi-Fi-based location estimation and navigation architecture which employs the existing IEEE 802.11 infrastructure for facilitating indoor positioning, providing business solutions, monitoring health care and guiding navigation. A statistical regression model is built on the recorded Received Signal Strength (RSS) dataset using Multi-output Least Square Support Vector Machine (M-LS-SVM) regression which infers the locality of a mobile device. The information from the radio map helps in improving the performance. The proposed M-LS-SVM technique is compared with various regression models for different kernels.