Anthropomorphic Features for On-Line Signatures.

Area of Science:

• Biometrics
• Human-Computer Interaction
• Computer Vision

Background:

• Online signature verification commonly uses positional and dynamic features from digital tablets.
• Existing methods may not fully capture the complex biomechanics of human handwriting.
• A need exists for more robust and informative features in signature verification systems.

Purpose of the Study:

• To propose a novel feature space for efficient online signature description.
• To leverage biomechanical principles by modeling the human arm's skeletal system.
• To enhance the accuracy and robustness of online signature verification.

Main Methods:

• Development of a virtual skeletal arm (VSA) model simulating human arm and forearm motion.
• Calculation of anthropomorphic features, including 3D joint positions and angles, from pen data.
• Utilizing VSA forward and inverse kinematic models to derive features from pen position and orientation.

Main Results:

• Demonstrated the effectiveness of the proposed anthropomorphic features.
• Achieved state-of-the-art performance with multiple verification algorithms.
• Validated robustness across diverse, third-party signature databases with varying collection methods and languages.

Conclusions:

• The novel anthropomorphic feature space significantly improves online signature verification.
• The VSA model provides a robust method for extracting biomechanically relevant features.
• This approach offers a promising direction for more secure and accurate biometric authentication.