Wireless LAN 802.11x has become a popular infrastructure in most public buildings. Many location system have been developed based on the wireless LAN. It is convenient for indoor location-aware services. However, recent research has demonstrated that localization error signi‾cantly increases in crowded and dynamic situations due to electromagnetic interferences. We propose collaborative localization to enhance the accuracy in the situation of human clusters by leveraging more accurate location information from nearby neighbors. We de‾ne "con‾dence" to represent the measurement error of each user. We use the Kalman ‾lter, interacting multiple model, and particle ‾lter to evaluate the con‾dence score. Our results showed a clear inverse relationship between con‾dence score and measurement error. Using con‾dence score as weight and neighborhood information, we can revise each user's measurement, and obtain the more accurate location estimation. Our experiments showed 28:2 - 56:0% accuracy improvement over the baseline system Ekahau, a commercial WiFi localization system.