Aiming at the problem that the navigation information source fusion algorithm of the U.S. military navigation sensor system interface relies on experience in parameter setting and does not fuse the course, a navigation information source fusion algorithm based on adaptive exponential weighted moving average (EWMA) is proposed. The algorithm uses the segmented estimation error function to adaptively adjust the parameters of the NSI algorithm, thereby avoiding the impact of improper initial parameter settings on the fusion accuracy, and further improving the position fusion accuracy. In addition, the algorithm is applied to the heading to realize the heading fusion. Under the simulation test and actual data test, the method in this paper is compared with the common NSI method based on EWMA. The results show that the NSI method based on adaptive EWMA proposed in this paper has the function of adaptive parameter adjustment, avoiding the complicated parameter adjustment process, and effectively fusing the information of multiple sensors to suppress the Strapdown Inertia Navigation System (SINS) error divergence.