Online Handwritten Signature Verification and Recognition Based on Dual-Tree Complex Wavelet Packet Transform

Area of Science:

• Biometrics and Security
• Signal Processing
• Machine Learning

Background:

• Technological advancements necessitate secure, user-friendly, and economical security systems.
• Handwriting signatures are a globally adopted biometric for ownership registration in finance and daily life.
• Online signature verification offers a powerful and publicly accepted method for personal authentication.

Purpose of the Study:

• To present a novel procedure for online signature verification and recognition.
• To enhance the accuracy and convenience of biometric security systems.

Main Methods:

• Utilized Dual-Tree Complex Wavelet Packet Transform (DT-CWPT) for three-level decomposition.
• Computed log energy entropy measures on DT-CWPT subbands to create feature vectors.
• Evaluated k-nearest neighbor, support vector machine, and Kolmogorov-Smirnov test classifiers.

Main Results:

• The DT-CWPT method was applied to dynamic signature data (position and pressure).
• Feature vectors were generated using log energy entropy measures.
• Classifiers were tested on Latin (SVC2004, MCYT-100) and Persian (NDSD) signature datasets.

Conclusions:

• Experimental results confirm the presented DT-CWPT method's effectiveness.
• The approach shows strong performance in both online signature verification and recognition tasks.
• This method offers a promising solution for advanced biometric authentication.