Optimizing 3D Face Recognition with PCA and CNN for Enhanced Accuracy

In the contemporary landscape of artificial intelligence, face recognition technology is proliferating across various domains, necessitating enhancements in accuracy and reliability. The prevalent 2D face recognition systems, despite their widespread application in security, are susceptible to breaches using photographs, highlighting a critical vulnerability. This paper proposes the transition to a 3D face recognition system, which inherently offers increased security through the utilization of depth data, a feature absent in two-dimensional images. The implementation employs a Time-of-Flight (ToF) camera, specifically the Kinect Xbox One, renowned for its reliability and cost-effectiveness. Previous endeavors in developing a 3D face recognition system with the Kinect Xbox One utilized Principal Component Analysis ( PCA) and Artificial Neural Networks (ANN), achieving an accuracy of approximately 80%. To surmount this limitation, this study integrates Convolutional Neural Networks (CNN), a more advanced algorithm, to refine accuracy. In this method, PCA reduces the dimensionality of the Kinect-captured images before processing them through CNN. Unlike standard resizing methods, which may lose critical details through uniform scaling, PCA focuses on the most significant features by transforming the data into orthogonal components. This combination improves computational efficiency and accuracy, achieving a notable increase in recognition accuracy to 90%. It is important to note that this research does not use a pre-collected dataset; instead, data is captured in real-time using the ToF camera, ensuring the most current and relevant data for evaluation. Consequently, this research substantiates the efficacy of combining PCA and CNN in enhancing the accuracy of 3D face recognition systems, thereby offering a more secure and dependable solution for system security applications.