Axial decoupled LS-SVMs for indoor positioning using RSS fingerprints

Location fingerprints based indoor positioning, which uses received signal strength (RSS) from wireless access points (APs), has become a hot research topic during the last a few years. Statistical learning based technique is one of the popular and effective methods for fingerprinting localization. Unfortunately, they suffer from high computational burden and require a large number of classifiers to determine the object location. To handle this problem, axial decoupled least squares support vector machines (AD-LS-SVM) based fingerprinting localization is proposed in this paper. First, the framework of fingerprinting localization based on AD-LS-SVM is given. Then, the decoupled training and positioning process by fingerprinting samples is described in detail. The attention is focused on how to transfer the positioning problem to a multi-class classification problem, for which we adopt two popular approaches: one-against-one (OAO) and one-against-all (OAA). Experimental results show that the proposed AD-LS-SVM method has the highest location accuracy among the traditional Grid LS-SVM, Support Vector Machines (SVMs) and the popular k-nearest neighbors (k-NNs), while requires much smaller training time.