Manuscript received August 11, 2025; revised September 30, 2025; accepted November 13, 2025; published April 28, 2026.

Abstract—Hand Gesture Recognition (HGR) is essential in enabling natural interaction between humans and machines. However, achieving high recognition accuracy remains challenging, especially when using lightweight machine learning models combined with classical feature extraction methods. This study evaluates the effectiveness of moment invariant features integrated with three classifiers (Support Vector Machine (SVM), Decision Tree (DT), and K-Nearest Neighbor (KNN)) for gesture classification. A dataset consisting of 1200 gesture samples is used to train and test the models. The evaluation results show that SVM achieved a near-perfect a testing accuracy of 81.41%. The DT classifier followes with a testing accuracy of 77.25%. Meanwhile, KNN obtaines a testing accuracy of 78.00%. These findings demonstrate that moment invariant features effectively capture discriminative geometric patterns in gesture data and highlight the superior performance of SVM for robust and efficient classification. The proposed approach offers strong potential for real-time gesture-based systems operating in resource-constrained environments, particularly where deep learning architectures may be impractical. This study contributes to the development of HGR solutions that are efficient, accessible, and suitable for deployment in low-power or limited-computation scenarios.

Keywords—decision tree, feature extraction, gesture-based interaction, human-machine interaction, k-nearest neighbor, pattern recognition, moment invariant, support vector machine

Cite: Fauziah*, Riadi M. Dinata, Dhieka A. Lantana, and Billy Hendrik, "Comparative Evaluation of Machine Learning Models for Hand Gesture Recognition Using Feature Extraction," Journal of Image and Graphics, Vol. 14, No. 2, pp. 312-324, 2026.

Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC-BY-4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.