A face and facial expression recognition method based on supervised isomap is proposed. The method can recognize the facial identity and facial expression simultaneously. Firstly, facial expression images were projected in low dimension manifold space. The points projected in the manifold space were separated according to their identity. At the same time, these images with the same identity were clustered in light of their expression class. Then, test images were embedded in the manifold space by using nonlinear mapping method. Finally, the identity and expression of test images were recognized by applying weighed k-nearest neighbor method. The experimental results in Cohn-Kanade and JAFFE face database show that the method can attain better recognition effect of face and facial expression.