Authentication Based on Pole-zero Models of Signature Velocity

Area of Science:

• Computer Science
• Biometrics
• Pattern Recognition

Background:

• Online signature verification is crucial for internet-based security, authenticity, and authorization.
• There is a need for fast and accurate algorithms in modern society.
• Biometric technology plays a significant role in access control.

Purpose of the Study:

• To model personal velocity signals for stable pattern recognition in signature verification.
• To develop a precise method for feature extraction from signature strokes.
• To evaluate the effectiveness of different classifiers for signature verification.

Main Methods:

• Utilized pole-zero models combined with Discrete Cosine Transform for signal modeling.
• Extracted features from signature strokes.
• Employed linear, Parzen window, and Support Vector Machine (SVM) classifiers.
• Tested the technique on authentic and forgery signatures from multiple databases.

Main Results:

• The proposed method demonstrated good potential for signature verification.
• Achieved equal error rates of 5.91% (Persian database), 5.62% (SVC2004), and 3.91% (SUSIG) on skilled forgeries.
• Effectiveness validated across diverse datasets, including proprietary, SVC2004, and Sabanci University signature databases.

Conclusions:

• The velocity signal modeling approach offers a precise and effective method for online signature verification.
• The technique shows promise for enhancing security in digital transactions and access control.
• Further validation across various databases confirms the robustness and potential of the proposed algorithms.