Personality Identification Based on Handwritten Signature

A individual has specific characteristics that form a unique personality that distinguishes that inividual from others and reflects the indivual's character and behavior when dealing with the environment. Graphology is the science of personality analysis based on hand-writing. Graphology expresses handwriting as a person's fingerprint in a written form. The signature is a mixture of writing and drawing and contains many indicators of an individual's charateristics. This paper addresses the challenge of a limited dataset in the field of graphology by collecting Arabic and English handwritten signature samples from students at King Abdulaziz University. Leveraging deep learning techniques, we extract signature indications from this diverse dataset, employing both single-label and multi-label classification approaches. The results demonstrate that single-label classification achieves 95% accuracy using ResNet and 91% using VGG16. Additionally, a custom convolutional neural network (CNN) was utilized, yielding an accuracy of 88%. In the case of multi-label classification, VGG16 and ResNet achieved an accuracy of 85%, while the EfficientNet model reached 86%.