A face-unlock screen saver by using face verification based on identity-specific subspaces

Abstract

A face-unlock screen saver prototype system is presented by using face verification technique based on identity-specific subspace. Instead of using password to unlock the screen saver, the proposed system can only accept the current legal user by his face as easily as just looking at the USB camera for a second, meanwhile refusing any other faces. The core techniques of the system are the proposed novel face verification algorithm based on identity-specific subspace and the maximum likelihood (ML) rule. Choosing these techniques is based on the fact that, for a task as screen saver, sufficient training facial images can be easily acquired. To achieve real-time verification, robust real-time facial feature localization and face normalization is presented, based on which an automatic samples collection tool is implemented. Practical experiments have shown the effectiveness of our proposed approaches.