Advancements in Gurmukhi Handwriting Analysis: Machine Learning Approaches for Attribute Classification

Handwriting analysis has broad applications in fields such as forensics, biometrics, and psychological profiling. This study investigates the classification of age, gender, and handedness from handwritten Punjabi documents using three machine learning approaches: k-Nearest Neighbors (KNN), Support Vector Machine (SVM), and a hybrid SVM-KNN model. A novel dataset comprising over 500 handwritten samples collected from 100+ individuals was created, with features extracted using advanced techniques including Histogram of Oriented Gradients (HOG), Speeded-Up Robust Features (SURF), Scale-Invariant Feature Transform (SIFT), Oriented Gradient Descriptors (OGD), and wavelet transforms. The models were evaluated based on classification accuracy, and results showed that while KNN provided a reliable baseline, SVM outperformed it by better capturing nonlinear relationships in the data. The hybrid SVM-KNN model achieved the highest overall performance, with accuracies of 79% for age, 80% for gender, and 78% for handedness classification, demonstrating its effectiveness in handling complex class overlaps. These findings underscore the importance of combining machine learning models to enhance predictive accuracy and establish a strong foundation for future studies in handwriting attribute analysis.