Face recognition based on variant 2DPCA-based approaches in robot environments

This paper is concerned with the performance comparison of variant two-dimensional principal component analysis (2DPCA) and its application to face recognition under robot environments. Face recognition is one of the most important components for human-robot interaction (HRI) technology that naturally interact between human and robot through robot camera, microphone, and various sensors. We demonstrate that the 2DPCA-based approaches show the insensitivity to large variation in illumination and distance variation in comparison to the conventional PCA and linear discriminant analysis (LDA). The comprehensive experiments are completed for ETRI face database. The robot platform used in this paper is WEVER, which is a network-based intelligent robot developed at Intelligent Robot Research Division in ETRI.