A Non-Parametric Kernel method for CDMA2000 network indoor localization using multiple observations

In this paper, we describe a Non-Parametric Kernel (NPK) method for localization of mobile users in the indoor CDMA2000 environment. This method makes use of CDMA2000 reference pilot, pilot delays, and pilot signal strength measurements to build a mobile signal fingerprint map in a given indoor environment in the off-line stage, and uses this fingerprint map to estimate the location of a mobile user in the on-line stage. The signal fingerprint map is in the form of a probability distribution function of the measurements created by the application of a kernel function to the off-line training data. The method is non-parametric in that no signal propagation model or parameter is used, and therefore avoids issues caused by the complicated signal propagation pattern often found in the indoor environment. We also propose the use of multiple consecutive observations to localize the user. Experiments are performed and the resulting Root Mean Square Error (RMSE) in localization accuracy demonstrates the advantage of our proposed method over other fingerprinting methods such as KNN.