Scholarlink

<p>Recognizing large pose variations in video-based facial recognition remains a major challenge due to factors such as occlusion, non-frontal angles, and illumination changes. These variations often reduce the accuracy and reliability of conventional recognition systems. To address this challenge, this paper introduces the Ensemble Dual-Weighted Kernel Extreme Learning-based Random Gazelle algorithm, which integrates robust feature extraction, ensemble voting, and metaheuristic optimization for effective pose variation recognition. The proposed model leverages the Dual-Weighted Kernel Extreme Learning Machine to capture complex facial patterns and employs ensemble voting to enhance classification accuracy. Furthermore, the Gazelle Optimization Algorithm with a random update strategy is applied for hyperparameter tuning, improving the system’s performance and adaptability. Experiments conducted on the Yoga Pose Videos and UCF101 datasets show that the proposed method achieves 98.84% accuracy, 97.83% precision, and an Area Under the Curve-Receiver Operating Characteristic score of 0.9874, outperforming several state-of-the-art methods. The results confirm the model’s robustness and efficiency in handling large pose variations, making it highly suitable for real-time facial recognition applications in dynamic environments.</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

Advanced signal, image, and video processing techniques for large pose variation detection with EDXL-RG

  • 期刊论文

  • 发表日期：

    2025年9月11日

查看全文
  • P. Jayabharathi,
  • A. Suresh

摘要

Recognizing large pose variations in video-based facial recognition remains a major challenge due to factors such as occlusion, non-frontal angles, and illumination changes. These variations often reduce the accuracy and reliability of conventional recognition systems. To address this challenge, this paper introduces the Ensemble Dual-Weighted Kernel Extreme Learning-based Random Gazelle algorithm, which integrates robust feature extraction, ensemble voting, and metaheuristic optimization for effective pose variation recognition. The proposed model leverages the Dual-Weighted Kernel Extreme Learning Machine to capture complex facial patterns and employs ensemble voting to enhance classification accuracy. Furthermore, the Gazelle Optimization Algorithm with a random update strategy is applied for hyperparameter tuning, improving the system’s performance and adaptability. Experiments conducted on the Yoga Pose Videos and UCF101 datasets show that the proposed method achieves 98.84% accuracy, 97.83% precision, and an Area Under the Curve-Receiver Operating Characteristic score of 0.9874, outperforming several state-of-the-art methods. The results confirm the model’s robustness and efficiency in handling large pose variations, making it highly suitable for real-time facial recognition applications in dynamic environments.