Hand shape classification is an important prob-lem for the human computer interaction and the fingerspelling recognition. For this matter, what is needed is a real-time processing and scale invari-ance. To this end, we propose a feature vector for hand shape classification, which is fast, and robust to scale. The proposed method calculates an adap-tive k-curvature which computes a curvature de-pending on hand size. The proposed method works at 0.1sec and has a 99% accuracy.