A face living body recognition method based on multi-classifier fusion of deep learning

Abstract

The invention discloses a human face living body recognition method based on multi-classifier fusion of deep learning, which comprises the following steps of respectively training three classifiers byusing traditional edge features, eye features and the like and deep neural network features, and then carrying out the human face living body recognition by using the classifiers through a designed program process. The method aims at a living body identification problem in a real scene. Frame detection, blink detection and moire detection are designed, the three detection results are fused innovatively according to a certain logic to obtain a human face living body recognition result, and the method has the advantages of being good in anti-cheating capacity, excellent in anti-interference capacity, capable of being matched by a small number of users, only needing a common camera (without additional equipment) and the like. The method mainly comprises the steps of the border detection, theblink detection, the moire detection, the fusion judgment and the like, and makes a certain contribution to the application of practical popularization of human face living body recognition.

Description

technical field [0001] The invention relates to scientific research fields such as machine learning, deep learning and pattern recognition, and in particular to a human face recognition method based on deep learning and multi-classifier fusion. Background technique [0002] Face liveness detection has become a key component of face recognition. It is a very important and necessary link for any biological detection system. It can ensure that the biological detection system can work safely and effectively; for unsupervised face For the application of recognition systems, automatically resisting photo and video spoofing is an urgent problem in the field of face recognition. [0003] The detection of live faces mainly includes the following methods: three-dimensional depth information analysis, optical flow estimation of facial motion, mixed face and voice recognition, Fourier spectrum analysis, eye blink detection, thermal infrared imaging recognition, etc. And the fusion of s...

AI Technical Summary

Problems solved by technology

In general, the current face recognition algorithm research focuses on feature extraction and classifier design. Most of the methods are not only unable to resist video attacks, but also require the assistance of additional equipment.

Images