Joint Bayesian Face Recognition Based on Data Adaptive Enhancement and Feature Fusion

In the face recognition, training data is sometimes limited, and even a single face photo per person, which causes the model less trained to recognize the multi-pose facial image with low recognition rate. In this paper, aiming at the face recognition using small sample training set, a joint Bayesian face recognition algorithm based on adaptive face data enhancement combined with global and local feature fusion is proposed. The face key points are detected and then expanded and rotated to extend the face data samples, and face model is established by combining the global and local features. The experiment result shows that the proposed algorithm makes the multi-gesture face recognition performance well.