Cellular network location estimation via RSS-based data clean enhanced scheme

Abstract

Cellular network based context-aware applications and location based services (LBSs) have drawn significant attention in both research and industry for many years. A key aspect that influences the quality of context-aware applications and LBSs is the localization accuracy of the mobile terminal (MT). The empirical location estimation method, also known as the fingerprint method, is a popular location estimation technique proposed for providing high accuracy position results. According to this method, the observed signals are compared with signals of known locations in the database. The closest location is determined to be the position of the unknown target. Most of the proposed algorithms for this method focus on predicting coordinate vectors of an unknown location using various prediction models and the empirical data. In this paper, we present a data clean scheme enhanced empirical learning algorithm (DCSEEL) which first minimizes the error that exists in the estimated distance between the target point and each reference base station (BTS). Then, the trilateration method [1] is used to calculate joints according to these distances. Unqualified joints will be excluded through a direction filter (DF). Synthetic experiment results confirm the superior performance of the proposed DCSEEL algorithm compare to the deterministic fingerprint techniques.