Pose Invariant Face Recognition System: Integrating PCA and SVM for improved identification

Abstract

Face recognition from online sources and other media has become a prominent research area. Advancements in face recognition and computer vision are continually sought in biometrics, banking, national identity systems, and law enforcement. However, many face recognition algorithms struggle with multi-constraint models. The field of computer vision and image processing is constantly evolving, with new research leading to modifications in existing techniques. Face recognition technology has progressed from basic methods to more sophisticated techniques and mathematical representations for face matching and image analysis. Achieving accurate face recognition with pose variations remains a challenge. This study explores face recognition and identification approaches using dimensionality reduction for pose-invariant datasets, examining techniques like PCA, LDA, and SVM. Computational complexity and time are crucial factors for accuracy and efficiency. This paper outlines the goals of face recognition applications and addresses associated complexities. PCA emerges as a strong dimensionality reduction technique, particularly when combined with SVM for classification. PCA reduces data dimensionality while preserving key features. With modifications, the algorithm can be designed for faster and more accurate face recognition, regardless of pose and color. While frontal face recognition is relatively straightforward, creating an accurate and efficient system for pose-invariant cases is a significant challenge