The reception state of a satellite is an unavailable information for Global Navigation Satellite System receivers. His knowledge or estimation can be used to evaluate the pseudorange. This article deals with the problem using three reception states: direct reception, alternate reception and blocked situation. This parameter, estimated using a Dirichlet distribution, is included in a particle filtering algorithm to improve the GNSS position in urban area. The algorithm takes into account two observation noise models depending on the reception of each satellite. Gaussian probability distribution is used with a direct path whereas a Gaussian mixture model is used in the alternate case.