Face recognition method based on stacked convolution sparse denoising auto-encoder

Abstract

The invention belongs to the field of face recognition, and particularly relates to a face recognition method based on a stacked convolution sparse denoising auto-encoder. The method comprises the following steps: alternately connecting a plurality of convolution sparse denoising auto-encoders with a plurality of pooling layers so as to construct a stack convolution sparse denoising auto-encoder model; removing the convolution form of the stack-type convolution sparse denoising encoder model, and training a stack-type sparse denoising auto-encoder model in a blocking mode; forming a convolution filter by using the trained parameters, and realizing a stacked convolution sparse denoising auto-encoder in a convolution form; performing face feature extraction by using a stack convolution sparse denoising auto-encoder, and classifying and recognizing faces by using a classifier. According to the invention, through the deep convolutional neural network which can be trained in an unsupervisedmanner, the face features having similar performance with the supervised deep convolutional neural network are extracted, and the training difficulty is effectively simplified.

Description

technical field [0001] The invention relates to a face recognition method, in particular to a face recognition method based on a stacked convolution sparse denoising self-encoder. Background technique [0002] As an artificial intelligence technology, face recognition has broad application prospects in the fields of national security, finance, and human-computer interaction. With the emergence of big data and high-performance computing, face recognition technology has developed rapidly, and face recognition under constrained environments (users are more cooperative and acquisition conditions are more ideal) has reached a practical level. However, in an unconstrained environment, it is difficult to achieve accurate face recognition because the face image is easily affected by uncontrollable factors such as illumination changes, expression changes, angle changes, age changes, and occlusions. In recent years, face recognition technology based on deep learning has greatly impro...

AI Technical Summary

Problems solved by technology

However, supervised learning models like DCNN require a large amount of labeled face data for training, and obtaining such data in reality is often costly and difficult to achieve

Images