Due to the inaccuracy of GPS devices, the location error of raw GPS points can be up to several hundred meters. Many applications using GPS-based vehicle location data require map-matching to pre-process GPS points by aligning them to a road network. However, existing map-matching algorithms can be limited in accuracy due to various factors including low sampling rates, abnormal GPS points, and dense road networks. In this paper, we propose the design and implementation of HIMM, an HMM-based Interactive Map-Matching system that produces accurate map-matching results through human interaction. The main idea is to involve human annotations in the matching process of some elaborately selected error-prone points and to let the system automatically adjust the matching of the remaining points. We use both real-world and synthetic datasets to evaluate the system. The results show that HIMM can significantly reduce human annotation costs comparing to the baseline methods.