Applying pattern recognition techniques based on hidden Markov models for vehicular position location in cellular networks

Abstract

Field trials of subscriber locations in a cellular network are discussed. The vehicular position location applied is a hybrid method based on pattern recognition and time of arrival (TOA) measurements. The pattern recognition is performed by hidden Markov models (HMMs) trained with prediction data to model the strength of the received signals for particular areas. The TOA gives first estimations of where the active mobile is located and which set of HMMs is to be used for the position estimation. To assess the accuracy of the proposed location method, calls have been performed from a car, driving through various streets and timing advance (TA) zones in a single GSM cell. The results are quite optimistic; the solution may fulfil the demand of many subscriber location applications, without requiring any modifications of existing standards, infrastructure or the mobiles.