In the Automatic Fingerprint Identification System (AFIS) , extracting the feature of fingerprint is very important .The local curvature of ridges of fingerprint is irregular , so people have the barrier to effectively extract the fingerprint curve features to describe fingerprint. This article proposes a novel algorithm; it embraces information of few nearby fingerprint ridges to extract a new characteristic which can describe the curvature feature of fingerprint. Experimental results show the algorithm is feasible , and the characteristics extracted by it can clearly show the inner macroscopic curve properties of fingerprint. The result also shows that this kind of characteristic is robust to noise and pollution.