This thesis presents the dynamic signature verification system using the shape and velocity characteristic function matching.
We improve the accuracy of the characteristic function comparison technique which is the hardest problem in the signature verification, by the modified regional correlation algorithm and elastic matching. Furthermore, the system determines the threshold value automatically and allows each user to have his own threshold to increase the performance.
The proposed system is composed of two major modules of the registration process and the verification process. The registration process enrolls user's reference signature and threshold into the signature database and the verification process discriminates the genuineness of the input signature. The preprocessing, segmentation, segment alignment, feature extraction and comparison procedures are included in both process modules.
In the performance evaluation, two error rates, type Ⅰ and type Ⅱ, are turned out low and the decision speed is fast. Particularly, using the velocity and the shape comparison methods, we can compare the genuine signature features with forged ones very well.
