Abstract: GPS is used for localization usually, but cannot be applied indoors. It was a popular topic on how to do the indoor localization accurately. In the early research, indoor localization system based on Wi-Fi fingerprints suffer from the accuracy and site survey problems. Therefore, a novel room-level indoor localization system was designed, which solved the localization problem by using multi-information fusion. Firstly, a top-n candidate through the Wi-Fi fingerprint was gotten, kNN classification algorithm was used after MAC address and the RSSI were gotten. The candidate set was filtered by the geomagnetic signal and the image information. Finally, the social information was used to give the final location results of people in the room. The system was validated on smart phone with Android system. The experimental results show that proposed method is more accurate than the Wi-Fi fingerprint localization algorithm.