Offline Handwritten Signature Analysis for Age Classification using Deep Features

Handwritten Signatures are prominent and convincing behavioral biometric data, used in many authentication application areas such as commercial, financial data, document analysis, health care, and forensic science etc. Different writing styles, cursive characters, and inconsistency in the shapes make it tougher to identify the writer from a handwritten signature. This problem remains still a challenging task, because of the considerable intra-class variations in handwriting. The proposed work focuses on the classification of the age using the different handwritten signatures collected from male and female writers of different age groups (adolescents (18-60)). In-house of total 6010 signatures were collected from 610 individuals. Experiments were carried out using well known deep neural architecture namely VGG16 model. Classification task carried out in end-to-end framework. The proposed work achieves comparatively better results than the existing methods respectively.