Velocity-image model for online signature verification

Area of Science:

• Biometrics
• Human-Computer Interaction
• Signal Processing

Background:

• Online signature capturing devices generate shape and velocity signals.
• Previous methods extracted strokes from velocity signal minimas, resulting in complex shapes that hinder template generation.
• This complexity impacts the discriminative capability of signature verification systems.

Purpose of the Study:

• To propose a novel stroke-based algorithm for online signature verification.
• To address the limitations of existing methods in generating discriminative signature templates.
• To enhance the accuracy and efficiency of signature verification through improved stroke extraction.

Main Methods:

• A new algorithm splits the velocity signal into various frequency bands.
• Strokes are extracted based on these segmented bands, resulting in simpler and smaller stroke representations.
• The medium-velocity band is identified as the most stable and discriminative for stroke extraction.
• Euclidean distances of strokes from the medium-velocity band are utilized for verification.

Main Results:

• The proposed algorithm extracts smaller and simpler strokes compared to traditional methods.
• The medium-velocity band proves effective for discrimination, while low- and high-velocity bands are found to be unstable.
• Experiments demonstrate an improved discriminative capability of the stroke-based system.
• The new method enhances the overall performance of online signature verification.

Conclusions:

• The proposed stroke-based algorithm offers a more effective approach to online signature verification.
• By utilizing a medium-velocity band for stroke extraction, the system achieves better discriminative power.
• This method simplifies stroke generation, leading to more robust and accurate signature templates.
• The findings suggest a significant improvement in the field of biometric authentication through enhanced signature analysis.