Face Recognition Method Based on Neural Network

Abstract

The invention relates to a face recognition method based on a neural network and belongs to the technical field of artificial intelligence and pattern recognition. The face recognition method is characterized by, at a training phase, firstly training the neural network through a known face image, achieving extraction of facial features through the learning process of the neural network, presenting descriptions of the facial features through the size of a connection weight, then testing the trained neural network by a training sample, and determining a classification threshold; and at a recognizing phase, inputting a face image to the recognized in the neural network, calculating a neural network output vector, and obtaining recognition results through comparing the largest component with the classification threshold. The face recognition method has the advantages that the method has high robustness to face image changes caused by various factors such as ambient lighting, viewing angles, expressions and making up; extraction and presentation of the 'similarity in spirit' feature in face images are achieved; and time for processing each face image is the same during recognition.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence and pattern recognition, and in particular relates to a face recognition method based on a neural network. Background technique [0002] In today's digital age, traditional identification methods based on keys, certificates, user names, and passwords can no longer meet the requirements for identity authentication. Identification methods that use the physiological and behavioral characteristics of people themselves, such as face, iris, etc. , fingerprints, etc. have attracted attention and become a research hotspot. The current face recognition method based on neural network mainly uses the neural network as a classifier. The input of the neural network is the shape, size, positional relationship and other characteristics of the key parts of the face. The neural network only completes the pattern classification according to these characteristics. It can be seen that these methods r...

AI Technical Summary

Problems solved by technology

It can be seen that these methods rely on some "seemingly effective features", which have problems such as subjectivity, large amount of calculation for acquiring features, and poor robustness.