In this paper, nonlinear Bayesian filtering techniques are applied to the localization of mobile radio communication devices. The application of this approach is demonstrated for the localization of DECT mobile telephones in a scenario with several base stations and a mobile handset. The received signal power, measured by the mobile
handsets, is related to their position by nonlinear measurement equations. These consist of a deterministic part, modeling the received signal power as a function of the position, and a stochastic part, describing model errors and measurement noise. Additionally, user models are considered, which express knowledge about the motion
of the user of the handset. The new Prior Density Splitting Mixture Estimator (PDSME), a Gaussian mixture filtering algorithm, significantly improves the localization quality compared to standard filtering techniques as the Extended Kalman Filter (EKF).