Dynamic Signature Verification System Based on One Real Signature

Area of Science:

• Biometrics
• Computer Science
• Pattern Recognition

Background:

• Automatic signature verification is crucial for identity authentication.
• Current systems require numerous signature samples, limiting practical application.
• Intrapersonal variability in signatures poses a challenge for accurate verification.

Purpose of the Study:

• To develop a signature verification system that trains effectively with a single reference signature.
• To reduce the data requirements for training biometric systems.
• To enhance the efficiency and accessibility of signature-based identity verification.

Main Methods:

• A novel procedure for duplicating a single reference signature using sigma lognormal decomposition.
• Two human-like signature duplication techniques: stroke-wise parameter variation and target-wise point modification.
• Training automatic signature verifiers with synthetically generated signature variations.

Main Results:

• The proposed system demonstrates robust performance on challenging benchmarks.
• Achieves comparable accuracy to traditional systems trained with multiple signature specimens.
• Effectively mitigates the need for extensive signature data collection.

Conclusions:

• A single reference signature is sufficient for accurate and robust identity verification.
• The sigma lognormal decomposition and duplication method significantly advances biometric system training.
• This research paves the way for more efficient and user-friendly biometric authentication solutions.