Handwritten Analysis for Gender Identification using CNN
O. Vignesh, R. Swathy, Sona Shiva · 2023
In this research, we provide a new method for determining a writer's gender based only on their handwriting samples, using deep learning methods. Our model utilizes a Convolutional Neural Network (CNN) to perform binary classification of the writer's gender and compares the performance of three different CNN architectures: ResNet50, VGG-19, and InceptionV3. The ICDAR 2013 dataset, which contains images of both English and Arabic handwriting samples, was used for training and testing the model. The unique features of the handwriting are discovered through CNN without the need for any prior feature extraction. The significance of this research lies in its application for forensic investigations, where demographic information such as gender, age, handedness, and ethnicity of the writer can be determined based on the handwriting left at a crime scene. This study demonstrates the potential of deep learning techniques in handwriting analysis and highlights the importance of developing models for gender prediction using only handwriting samples.