Indoor positioning techniques based on radio fingerprints outstand over other localization methods because of their independence from radio propagation models and cost-effectiveness in terms of hardware and deployment requirements. However, their reported best achieved accuracy is bounded due to the random environmental changes which cause the inconsistency between the stored fingerprints and the current radio behavior. In order to overcome this limitation, we propose a cooperative localization scheme, whereby users exchange their real-time signal measurements in order to update and improve their estimated location. The update process relies on a modified version of the neural network structure of Self-Organizing Maps by considering the signal relationship between users. Performance evaluation results demonstrate accuracy improvement over the baseline fingerprinting technique while keeping the communication and complexity overheads low