A Comparative Study of Face Recognition Techniques

Abstract Face recognition is one of the most widely used biometric methods for identifying or verifying a person from a digital image or video frame. Over the last few decades, a number of techniques have been developed for this task, ranging from early statistical approaches such as Eigenfaces and Fisherfaces to modern deep learning based methods such as FaceNet and ArcFace. This paper presents a comparative study of these techniques by discussing how each method works, along with their advantages and limitations. A simple comparison of these methods is also presented based on accuracy, robustness and computational cost, using values reported in earlier studies. The paper also discusses some common challenges faced in face recognition, such as changes in lighting, pose and occlusion, and lists some of the areas where this technology is applied today. The study shows that while traditional methods are simple and fast, deep learning based methods give much higher accuracy and are more suitable for real world conditions. Keywords: Face Recognition, Eigenfaces, LBPH, Convolutional Neural Network, FaceNet, ArcFace, Biometrics