Siddalingappa, Rashmi ORCID: https://orcid.org/0000-0001-9786-8436, Gornale, Shivanand, Kumar, Sathish and Mane, Arjun (2022) Gender Classification Based on Online Signature Features using Machine Learning Techniques. International Journal of Intelligent Systems and Applications in Engineering (IJISAE), 10 (2). pp. 260-268.

Preview

Text Gender_Classification.pdf - Published Version Available under License Creative Commons Attribution Share Alike. Download (410kB) | Preview

Abstract

A human signature gives a lot of insights into an individual’s characteristics such as illness, professional choices, and mood. From the biometric perspective, a Handwritten Signature is a behavioral trait and Gender is a demographic category (soft trait) of the person. Gender classification from handwritten signatures has been implied in several applications such as psychology and forensics. Male writings with a high intra-class variation tend to have a feminist aesthetic aspect, and vice versa. This gives clues to recognize the gender of the person using a handwritten signature. The proposed methodology is based on extracting numeric features from the male and female dynamic signature samples. Data was collected from 535 individuals of different age groups (18-65). Further, these signature samples were converted to numeric attributes resulting in 66 signature features from each data. Experiments were carried out using six different Machine Learning techniques; On the whole, the overall accuracy of these methods is 81.2% (KNN), 81.9% (LR), 77.1% and 49.3% (for both Poly and RBF kernels in SVM, respectively), Poly kernel using cross-validation resulted in 81.8% in SVM, 89.3% (DT), 96.2% (RF) and 98.2% (DL). Overall, the deep neural networks outperformed other models, immediately followed by RF.

Item Type:	Article
Status:	Published
Subjects:	Q Science > Q Science (General)
School/Department:	London Campus
URI:	https://ray.yorksj.ac.uk/id/eprint/13306

University Staff: Request a correction | RaY Editors: Update this record