Accurate and Low-cost Location Estimation Using Kernels

Abstract

We present a novel method for indoor-location estimation using a vector-space model based on signals received from a wireless client. Our aim is to obtain an accurate mapping between the signal space and the physical space without incurring too much human calibration effort. This problem has traditionally been tackled through probabilistic models trained on manually labeled data, which are expensive to obtain. In this paper, we present a novel approach to building a mapping between the signalvector space and the physical location space using kernel canonical correlation analysis (KCCA). Its training requires much less human labor. Moreover, unlike traditional location-estimation systems that treat grid points as independent and discrete target classes during training, we use the physical location as a continuous feedback to build a similarity mapping using KCCA. We test our algorithm in a 802.11 wireless LAN environment, and demonstrate the advantage of our method in both accuracy and its ability to utilize a much smaller set of labeled training data than previous methods.