Regional Features Selected Based on GAs in Online Signature Verification
Mengtian Cui, Yong Zhong, Haijun Zhao · 2008
A new method to select a segment-to-segment matching by analyzing signature verification is presented, accordingly curve segments used in signature verification and the regional feature contained in the curve segment are picked-up and the regional features are selected by GAs. Namely, features selected are first encoded into chromosome, and descendible types are founded by GAs improved locally. A new crossover method is also proposed to determine the number of curve segment. The experiment shows that the algorithms proposed can accurately find optimal features for signature verification and bring the lower FRR and FAR, thereby the veracity in online signature verification is enhanced.