Method and device for face authentication

Abstract

The invention discloses a method and device for face authentication, belonging to the field of biometrics. The method includes: using a multi-level deep convolutional network that has been jointly trained by a multi-layer classification network in advance for a face image and a face image template to be authenticated Extract the feature vectors of multiple levels in sequence; map the feature vectors of multiple levels into uniform dimension feature vectors through the uniform dimension linear mapping matrix in turn; concatenate the uniform dimension feature vectors into joint feature vectors; reduce the joint feature vector through linear dimensionality reduction The dimensionality reduction mapping of the mapping matrix is ​​performed to obtain the comprehensive feature vector; through linear discriminant analysis, the absolute value is used to normalize the cosine value, and the obtained comprehensive feature vector of the face image to be authenticated is compared with the comprehensive feature vector of the face image template certified. Compared with the prior art, the face authentication method of the present invention has strong anti-interference ability, good scalability and high authentication accuracy.

Description

technical field [0001] The invention relates to the field of biometric identification, in particular to a method and device for face authentication. Background technique [0002] Face authentication is a form of biometric identification. By effectively characterizing the face, the features of two face pictures are obtained, and a classification algorithm is used to determine whether the two pictures are the same person. Generally, a face image is pre-stored in the face recognition device as a face image template; during authentication, a face image is taken as a face image to be authenticated, the features of the two images are extracted, and classification algorithms are used to Determine if the two photos are of the same person. [0003] The method of feature extraction is: manually design a feature vector, and extract the specified feature vector through various algorithms, such as face authentication methods based on geometric features, face authentication methods based...

AI Technical Summary

Problems solved by technology

[0003] The method of feature extraction is: manually design a feature vector, and extract the specified feature vector through various algorithms, such as face authentication methods based on geometric features, face authentication methods based on subspace, and face authentication methods based on signal processing. etc., but this method is extremely susceptible to the influence of lighting, expression and other factors on the results, and has poor anti-interference ability, and the artificially designed feature vectors are mostly based on specific situations, and the scalability is poor.

[0004] Face recognition and authentication technology based on deep network can automatically learn and extract features, but the general deep network has the problem of gradient dispersion, and the processing and understanding of features at each level is insufficient, and only using high-level features is not enough to fully describe the image

Images