Evaluation of the Off-line Signature Verification by the Hybrid Model

摘要

In this paper, we present an off-line signature verification method which combines three different algorithms for improving precision. Firstly, we got average error rate 19.0[%] as a result of DP matching with Gaussian filter, and 20.75[%] with thinning. A principal analysis method got an average error rate 23.5[%] in case of dividing to nine regions. We propose the system which extracts the individual characteristics of signatures by thinning and the Gaussian filtering. Furthermore, we combine it with legacy principal component analysis methods. We got 9.0[%] as a result of proposed method.