A Method of On-Line Signature Verification with Consistent Spectral Features
Xinghua Xia, Fangjun Luan, Zhili Chen · Advances in intelligent systems research/Advances in Intelligent Systems Research · 2015
Consistency analysis of signature features has significant effect on the performance and robustness of online signature verification.Inorder to improve the performance of on-line signature verification, we propose a method of on-line signature verification based on spectral consistent features, in which the consistent and effective information is extracted of signature features to be used in verification.In our works, spectral consistency of features is analyzed by statistics along with the consistent penalty factor.The consistent spectral features are obtained dynamically by the wavelet packet reconstruction with consistent sub-bands.Similarity between test signature and reference is calculated by proposed CPF-DTW to improve the robustness.Several experiments are carried out both on CDB and DB1 which consists of 8150 signatures from 190 individuals in total, the best results of our proposed method are given by EERCDB=2.25%and EERDB1=2.38%respectively which indicate the effectiveness and robustness of our proposed methods.