In order to resolve the shortcoming of the federated Kalman filter, the combination of the standard Kalman filter method and the confidence weighted method is applied in the integrated navigation system. In the method the advantages of Adaptive-Network-based Fuzzy Inference System: non-linear; fast and real-time; adaptive learning are used to compute the confidence of filters' output data, then get the weighting factor of subsystems, and then fuse with these weighting factors to compute the comprehensive outputs. The simulation result shows that the new algorithm prevents the divergence of data effectively and the precision of navigation is improved.