Accurate localization of mobile phone users is needed for search and rescue applications or for electronic surveillance in urban or suburban areas. Assuming that the mobile station (MS) has been roughly localized by the cellular network or by any other means, we propose an approach using dedicated base stations (BSs), deployed in the zone where the MS is situated and forcing the MS to transmit. In this work, a fingerprinting framework is proposed to localize the MS in this non-cooperative mode and outdoor non-line-of-sight (NLOS) conditions. First, the wireless channels between the MS and one or several BSs are measured to estimate the angular and delay characteristics of the most powerful rays. In a second step, the multipath component distance metric is used to compare the channel characteristics to those predicted by a ray-launching model for all possible MS positions within the area of interest. After describing in detail the principle of the method, a statistical analysis of the location errors is presented based on the simulation of both the MS-BS and the BS-MS links.