A real-time multi-user face unlock system via fast sparse coding approximation

We propose a multi-user face unlock system based on fast sparse coding approximation. Different approximation techniques of sparse coding are compared for real-time processing and recognition rate. The system is capable of online new user registration without specific training, and it is designed for continuously processing video sequences as on-the-fly face classification. The system is tested with a public face dataset for real-world applications, which is able to recognize 100 different face identities in 0.0028s on a regular PC, with 93.8% recognition rate.