Indian Sign Language Alphabet Recognition Using a Hybrid CNN-PCA-PSO-SVM Framework
International Journal of Innovative Science and Research Technology · 27.04.2026 · 1954
DOI: 10.38124/ijisrt/26apr960 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Innovative Science and Research Technology, online erschienen am 27.04.2026. Verfasst von Girish Katkar und 2 weiteren Autor:innen. Die Arbeit nennt 9 Literaturangaben.
Abstract (Originalsprache)
In this work, we propose an efficient hybrid framework for recognizing Indian Sign Language (ISL) alphabets by combining deep feature extraction with classical machine learning and optimization techniques. A pre- trained MobileNetV2 network is utilized to extract discriminative visual features from hand gesture images. These highdimensional features are subsequently compressed using Principal Component Analysis (PCA) to eliminate redundancy and improve computational efficiency. Particle Swarm Optimization (PSO) is then employed to determine optimal hyperparameters for a Support Vector Machine (SVM) classifier. The proposed system is evaluated on a 26-class ISL dataset and achieves an overall accuracy of 96.17%, along with consistently high precision, recall, and F1-scores. Further validation using confusion matrix and ROC analysis demonstrates the robustness and strong class separability of the model.
Zitieren
APA
Katkar, G., Gaikwad, S., & Ramteke, A. (2026). Indian Sign Language Alphabet Recognition Using a Hybrid CNN-PCA-PSO-SVM Framework. International Journal of Innovative Science and Research Technology, 1954. https://doi.org/10.38124/ijisrt/26apr960
Vancouver
Katkar G, Gaikwad S, Ramteke A. Indian Sign Language Alphabet Recognition Using a Hybrid CNN-PCA-PSO-SVM Framework. International Journal of Innovative Science and Research Technology. 2026:1954. doi:10.38124/ijisrt/26apr960
BibTeX
@article{Katkar2026,
author = {Katkar, Girish and Gaikwad, Shalaka and Ramteke, Ajay},
title = {Indian Sign Language Alphabet Recognition Using a Hybrid CNN-PCA-PSO-SVM Framework},
journal = {International Journal of Innovative Science and Research Technology},
year = {2026},
pages = {1954},
doi = {10.38124/ijisrt/26apr960}
}
EUVIMED