Pedestrian navigation services guide people to reach their destinations as the vehicle navigation services do. However, the moving way of people differs from that of vehicles, and hence the assumptions for car-navigation services are not suitable for pedestrian navigation. People may walk through the place without GPS signals due to the shelter in the sky. In some space, the accuracy of GPS may not be enough for pedestrian navigation. In this situation, localization ability is necessary in this space, we call it as a special interest zone (SIZ), to assist the GPS-navigation systems. This paper proposes a system by offering a navigation service on general routes and a localization service to SIZ. For navigation service, GPS and GIS technologies were used for guiding, and a modified A* algorithm was developed to implement the path planning function. Because the accuracy of GPS is insufficient to offer the precise localization needed at SIZ, ZigBee-based sensor networks were applied and deployed around SIZ. When the user is close to SIZ, a dynamic swap mechanism enables the localization function to provide higher localization results. The localization algorithm, using an extended Kalman filter to fuse the data of ZigBee and GPS, achieved the localization error less than 1 m, when the error compared to ZigBee-only or GPS-only approach. Both services were successfully implemented on an embedded platform and evaluated on the real environment.