The performance of localization techniques in a wireless communication system is severely impaired by biases induced in the range and angle measures because of the non-line-of-sight (NLOS) situation, caused by obstacles in the transmitted signal path. However, the knowledge of the line-of-sight (LOS) or NLOS situation for each measure can improve the final accuracy. This paper studies the localization of mobile terminals (MT) based on a Bayesian model for the LOS-NLOS evolution. This Bayesian model does not require having a minimum number of LOS measures at each acquisition. A tracking strategy based on a particle filter (PF) and an unscented Kalman filter (UKF) is used both to estimate the LOS-NLOS situation and the MT kinetic variables (position and speed). The approach shows a remarkable reduction in positioning error and a high degree of scalability in terms of performance versus complexity.