A priori selection of high accuracy mobile station position estimates

A priori identification and selection of high accuracy position estimates, i.e., with error below 100 meters, is particularly relevant for critical location-based applications, like vehicle tracking and, specially, emergency call positioning. This work presents a backpropagation artificial neural network classifier used to predict the accuracy of mobile station position estimates produced by a network-based radio-frequency fingerprinting method, RF-FING+RTD-PRED (Predicted Radio-frequency Fingerprint with Round Trip Delay), previously formulated by the authors. The classifier employs the same radio-frequency parameters used by the aforementioned method plus some additional network data. In field tests carried out in GSM (Global System for Mobile Communications) networks in urban and suburban areas, where 6600 measurement reports have been collected, a 89% precision in the identification of high accuracy position estimates has been achieved. The presented method is promptly extensible to 3G cellular networks.