COMPARATIVE ANALYSIS OF HANDWRITTEN CHARACTER RECOGNITION USING ML MODELS
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Malaysia Sarawak (UNIMAS)
Abstract
Description
Handwritten Character Recognition (HCR) remains a pivotal research domain within Machine Learning (ML), intersecting computer vision, pattern recognition and human-computer interaction. Its applications spread as critical areas such as digitizing historical manuscripts, automated postal sorting, bank check processing, and assistive technologies. While numerous studies have explored HCR using diverse ML methodologies deep learning architectures (e.g., CNNs, RNNs, Transformers), and hybrid approaches. Each model presents distinct strengths and limitations in terms of accuracy, computational efficiency, robustness to variability (script, style, noise), and scalability. This study conducts a comprehensive comparative analysis of prominent ML-based HCR models, employing rigorous benchmarking across standardized datasets and diverse scripts. Performance is evaluated quantitatively (accuracy, precision, recall and F1-score) and qualitatively (error pattern analysis). As the result, the Character Error Rate of each model is as follow CNN(49.76%), CNN-LSTM (43.10%) and TrOCR (36.24%)