Scholarlink

<p>Gender classification through handwriting analysis presents significant challenges due to subtle similarities in writing styles between males and females. This study explores handwriting as a behavioural biometric trait, specifically focusing on the Gurmukhi script—a script that has been relatively underexplored for gender classification. The research integrates multiple feature sets and employs Principal Component Analysis (PCA) for dimensionality reduction, aiming to enhance the performance of machine learning algorithms. Feature extraction techniques such as zoning, diagonal, transition, and peak extent-based methods were utilized. Decision Tree and Random Forest classifiers were applied, achieving maximum accuracies of 81.48% and 89.85%, respectively, without PCA. When PCA was applied, the accuracy of the Decision Tree improved to 88.17%, while the Random Forest accuracy remained unchanged at 89.85%. The highest accuracy achieved with Random Forest was 90.21% using the feature combination of F1 + F2 + F3 + F4, and the lowest False Positive Rate (FPR) of 0.19% was obtained using F2 + F3 + F4 with Random Forest. Future directions for this work include investigating classification approaches without pre-segmentation, enabling online gender recognition, and incorporating transgender handwriting samples to improve the inclusivity of the model. This study contributes to the advancement of handwriting-based biometric authentication systems for gender classification.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Gender Classification Through Handwriting Analysis of Gurmukhi Script Using Ensemble Feature Extraction and PCA

  • 期刊论文

  • 发表日期：

    2026年5月24日

查看全文
  • Ahmed Alkhayyat,
  • Junainah Abd Hamid,
  • B. Jayaprakash,
  • Krishan Kumar Sah,
  • Mandeep Kaur Chohan,
  • Amrita Singh,
  • S. Srinadh Raju,
  • Rajesh Singh

摘要

Gender classification through handwriting analysis presents significant challenges due to subtle similarities in writing styles between males and females. This study explores handwriting as a behavioural biometric trait, specifically focusing on the Gurmukhi script—a script that has been relatively underexplored for gender classification. The research integrates multiple feature sets and employs Principal Component Analysis (PCA) for dimensionality reduction, aiming to enhance the performance of machine learning algorithms. Feature extraction techniques such as zoning, diagonal, transition, and peak extent-based methods were utilized. Decision Tree and Random Forest classifiers were applied, achieving maximum accuracies of 81.48% and 89.85%, respectively, without PCA. When PCA was applied, the accuracy of the Decision Tree improved to 88.17%, while the Random Forest accuracy remained unchanged at 89.85%. The highest accuracy achieved with Random Forest was 90.21% using the feature combination of F1 + F2 + F3 + F4, and the lowest False Positive Rate (FPR) of 0.19% was obtained using F2 + F3 + F4 with Random Forest. Future directions for this work include investigating classification approaches without pre-segmentation, enabling online gender recognition, and incorporating transgender handwriting samples to improve the inclusivity of the model. This study contributes to the advancement of handwriting-based biometric authentication systems for gender classification.