Related Experiment Video
Updated: Aug 16, 2025
06:37
Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
Published on: December 15, 2023
4.0K
Automatic Gender and Age Classification from Offline Handwriting with Bilinear ResNet
Irina Rabaev1, Izadeen Alkoran1, Odai Wattad1
1Software Engineering Department, Shamoon College of Engineering, 56 Bialik St., Be'er Sheva 8410802, Israel.
Sensors (Basel, Switzerland)
|December 23, 2022
Summary
This study introduces B-ResNet, a novel deep learning model for automatic gender and age prediction from handwritten documents. B-ResNet achieves top performance across multiple languages, advancing writer demographics classification.
Area of Science:
- Computational linguistics
- Forensic science
- Digital humanities
Background:
- Automatic gender and age prediction from handwriting is crucial for historical document analysis and forensic investigations.
- Existing methods show limited performance, and deep neural networks have not been applied to age classification.
- Research has primarily focused on English and Arabic, neglecting other languages like Hebrew.
Purpose of the Study:
- To develop and evaluate a novel deep learning approach for automatic gender and age classification from handwritten documents.
- To apply bilinear Convolutional Neural Network (B-CNN) with ResNet blocks (B-ResNet) for writer demographics classification.
- To investigate the model's performance on multiple languages, including under-researched ones like Hebrew.
Main Methods:
- Implementation of a novel bilinear Convolutional Neural Network (B-CNN) integrated with ResNet blocks, termed B-ResNet.
- Application of the B-ResNet model to gender and age classification tasks on handwritten documents.
- Experimental evaluation on three benchmark datasets covering English, Arabic, and Hebrew languages.
Main Results:
- The proposed B-ResNet model achieved top-ranked performance across all evaluated tasks (gender and age prediction).
- B-ResNet demonstrated superior performance compared to existing models, particularly on the KHATT and QUWI datasets for gender classification.
- This work represents the first application of B-CNN for writer demographics and the first deep neural network approach for age classification in this context.
Conclusions:
- The B-ResNet model offers a significant advancement in automatic gender and age prediction from handwritten texts.
- The model's effectiveness across different languages highlights its potential for broader applications in historical and forensic document analysis.
- This research establishes a new benchmark for deep learning-based writer demographics classification.
