Human Face Recognition Method Based On Neural Network Research

Face picture recognition is one of research focuses in the fields of the pattern recognition and artificial intelligence currently. By researching and analyzing the domestic and overseas research status of face picture recognition, the article discussed face picture recognition method based on neural networks.This paper studies face picture recognition method based on neural networks, and under the experiment realizes face location based on the skin color characteristic; to the face location, the image after being detected is projected respectively to level and vertical direction; we obtain the good result and increase to the efficiency of feature distilling through using the method of geometrical feature distilling to make accurately face location; in the end, face pictures are identified through the method based on RBF neural networks. To the recognition of small sample, the correctness of recognition can exceed 80%, if the parameters and initial values are correctly selected. A new approach to face recognition combining geometry feature distilling and neural networks methods is proposed. Advantages of the approach are: Don't require face images to be peculiarly pretreated, that face images were enhanced or not didn't affect recognition rate; it was not sensitive on face gesture, expression and illumination.This paper studies face picture recognition method , but the following problems are existing: to match actual application request, the efficiency of arithmetic should be improved; the face pictures which have color deviation are difficultly detected; face feature of choice and distilling arithmetic should be optimized; the effect is not ideal to recognition of the two images of same person, which are different from in the facial expression; the collection of the face pictures is not enough extensive. To achieve the actual applied request, we need to make the further development in above aspects.