Abstract

Off-line handwritten biometric recognitionbased on stroke direction distribution features

Author(s): Rengui Cheng, Hong Ding, Xiaofeng Zhang

Off-line handwritten biometric recognition (OLHBR) is an authentication method based on writing features detected from differenthandwriting images. It remains a challenge for no dynamicwriting order can be used. In this paper, a method based on stroke direction distribution feature (SDDF) is proposed for OLHBR. The proposed feature is utilized for catchingthe distribution features of strokes, which are counted by a loop counting procedure. Then, the similarities between features are measured by the weighted Manhattan distance. In order to reduce the impact of the stroke thickness, two methods have been applied in our method. One is decomposing the whole contour into strokes. Another is ignoring the fragments not connecting the center pixel in a current loop counting window. At last, the experiments on ICDAR 2011writer identification database show the effectiveness of the proposed method.