A new online signature-verification method using temporal shift estimation is presented. Local temporal shifts existing in signatures are estimated by the differences of the phase outputs of Gabor filter applied to signature signals. In the proposed signature-verification algorithm, an input signature signal undergoes preprocessing procedures including smoothing, size normalization and skew correction, and then its feature profile is extracted from the signature signal. A Gabor filter with the predetermined center frequency ω is applied on a feature profile, and a phase profile is computed from the phase output. The feature profile and the phase profile are length normalized and quantized so that a signature code of fixed size is generated. The temporal shifts existing between two signatures are computed by using the differences between the phase profiles. The information about the temporal shifts is used as offsets for comparing the two feature profiles. Therefore, two kinds of dissimilarities are proposed. Temporal dissimilarity is a measure reflecting the amount of total temporal disturbance between the two signatures. The difference between the two signature profiles is computed at each corresponding point pair and is accumulated into temporally arranged feature profile dissimilarity. The decision boundary is represented as a straight line in the dissimilarity space whose two axes are the two dissimilarity measures. The slope and the position of the decision boundary are computed using the distribution of the dissimilarities among the sample signatures involved in the enrollment procedure. The experimental results show that through the compact and fixed size of signature data representation and relatively simple comparison methods, the proposed method can compare signatures 400 times faster than the conventional DP matching based signature-verification method.
Bibliographical noteFunding Information:
Manuscript received October 29, 2003; revised July 13, 2004. The associate editor coordinating the review of this paper and approving it for publication was Dr. Stefan Katzenbeisser. This work was supported by Korea Science and Engineering Foundation (KOSEF) through the Biometrics Engineering Research Center, Yonsei University.
All Science Journal Classification (ASJC) codes
- Signal Processing
- Electrical and Electronic Engineering