We propose a fingerprint verification algorithm with template matching based on the cyclic structure of the fingerprint image. For a correlation value between a template and a sensed image, we use the differential matching rate. The differential matching rate utilize the cyclic structure observed in the local area of a fingerprint pattern, which is calculated by the maximum matching rate minus the minimum matching rate detected near the point where the matching rate is maximum. In order to extract the maximum effectiveness of the cyclic structure, we compare divided windows of a template image against a sensed image. A verification experiment confirmed that the proposed method judges whether a template image and a sensed image are same finger?s fingerprints more accurately than the conventional method using the maximum matching rate as a correlation value. We obtain the equal error rate (EER) of 3.6%, which is better than half that of the conventional method.