A METHOD OF FACE RECOGNITION USING PCA AND BACK-PROPAGATION ALGORITHMS

摘要

The present invention provides a face recognition method, based on the expression of the face image vector, that is, extracting eigenvectors After the learning by using each of the vector space is defined as the characteristic value for which the projection on the weight PCA (Principal Component Analysis) and before performing a recognition by the face image and the extracted feature value, to perform the recognition relates to a back propagation (Back-Propagation) algorithm using a face recognition method. ; More specifically, the present invention is the appropriate size for the process of recognition and to be less sensitive to the illumination takes a face image pre-treatment step to convert; Calculating a weighted PCA process to extract a set of a specific face of the rough face data process; Thus obtained through a learning process to update outgoing to a certain error range weight for each of the data inverse spread processing step; And face recognition step using standard deviation; made, including, ; After extracting the facial image of him as a genus-specific face PCA to reduce the feature vector of the vast amount of input when using BP neural network only relates to a face recognition method using the PCA and input to the neural network back-propagation algorithm. ; According to the present invention, while reducing the dimension of the original by using the eigenvector obtained by applying PCA to the original face image On the other hand represents well the features of the face, it is possible to shorten the time to learn the use of the neural network type.