Hybrid Adaptive Face Recognition Framework (HAFRF): A Machine Learning Approach Integrating YOLOv11, FaceNet, Adaptive PCA, and Support Vector Machine

Abstract

The face recognition technology has gradually evolved into an essential one in the field of Intelligent Surveillance, biometric authentication, Access Control and HCI. Most of the traditional face recognition systems suffer from low performance due to illumination changes, facial expressions, pose and occlusion issues in real-world applications. To overcome these challenges, this paper proposes a machine learning-based method called Hybrid Adaptive Face Recognition Framework (HAFRF) that combines the following four modules: (1) YOLOv11 for accurate face detection, (2) FaceNet for discriminative facial feature embedding, (3) Adaptive Principal Component Analysis (Adaptive PCA) for efficient dimensionality reduction, and (4) support vector machine (SVM) for accurate identity recognition. The proposed framework incorporates strong feature extraction along with adaptive feature optimization, which will enhance the recognition performance with minimum computational complexity. The effectiveness of the proposed framework is tested in various facial conditions using publicly available facial image datasets. Experimental analysis shows that HAFRF is superior to the conventional machine learning techniques in terms of recognition accuracy, precision, recall, F1_score and processing time. By combining deep feature representation with adaptive dimensionality reduction, the framework's ability to recognize faces in unconstrained environments is enhanced. The proposed HAFRF is scalable and reliable for next generation biometric authentication system and can be applied in security, surveillance and smart access control applications.