Human face recognition is one of the relative difficult and important issues in the field of pattern recognition, which is significant both in theoretical research and in practical application. Traditional methods of face recognition based on K-L transform do not necessarily correspond to isolated features such as eyes, ears, and noses. And face recognition using eigenfaces has made much progress. But for complex and special scenes, facial images will change as a result of the effect of age, emotion, illumination changes, and hair, so the images of the same person differ drastically. Traditional methods of automatic face recognition based on K-L transform can not cope with such effects. In this paper, principal component analysis is introduced into face recognition. Simulating the real changes of facial images and making corresponding distortions in advance, a series of distorted facial images are produced. Then the principal components of distorted faces are extracted with principal component analysis. Finally appropriate principal components are selected with genetic algorithm to span an eigenspace for recognition.A new distortion invariant method of automatic face recognition with learning function is proposed and experiments with this method are made.Experimental results show the feasibility and robustness of this algorithm.