Although face recognition has been applied to a wide variety of problems, ranging from security to virtual reality systems; such as many researchers pay more attention in it; the complexity of the algorithm used in face verification is not often suitable for real-time face recognition. This paper proposes an excellently efficient feature extraction method, combining with probabilistic neural network (PNN) for real-time face recognition. The proposed method is evaluated on the ORL face database. The experimental results show that the proposed method possesses excellent recognition performance and efficiency. Finally, the face authentication recognition system is implemented on a PC and the best recognition rate is 100% in real environment.