Face micro-expression deep learning recognition system based on combined adversarial generative network

Abstract

The invention provides a face micro-expression deep learning recognition system based on a combined adversarial generative network, which is used for training a model for face micro-expression recognition so as to complete face micro-expression recognition of a to-be-recognized image, and is characterized by comprising a model storage part, wherein the model storage part stores a multifunctional recognition network which is trained in advance and is used for carrying out micro-expression recognition, posture classification and face recognition and a micro-expression face image generation network based on a combined adversarial generation network; a to-be-recognized image acquisition part used for acquiring a to-be-recognized image; a feature extraction part used for inputting the to-be-recognized image into the multifunctional recognition network so as to obtain feature information corresponding to the identity, the posture and the micro-expression; and a micro-expression recognition part used for completing face micro-expression recognition according to the feature information.

Description

technical field [0001] The invention belongs to the technical field of computer image recognition, and in particular relates to a human face micro-expression deep learning recognition system based on combined confrontation generation network. Background technique [0002] Micro-expression is an effective message to convey the true emotions and psychology of human beings, and it is an effective expression of personal intentions. Since Darwin's "Expressions of Humans and Animals" explained the universality and continuity of facial expressions in humans and animals, and our microexpressions will reflect much more information than we imagined, microexpression research It is carried out in many fields, such as clinical medicine, public safety, education and political fields. [0003] Therefore, for the task of automatic recognition of facial micro-expressions, due to the difficulty of observing micro-expressions with the naked eye, some micro-expression automatic recognition met...

AI Technical Summary

Problems solved by technology

[0006] Literature [1] (Wang, Zhanxiong, et al. "Multi-task deep neural network for joint face recognition and facial attribute prediction." Proceedings of the 2017ACM on International Conference on Multimedia Retrieval. ACM, 2017.) proposes the facial expression recognition model Pre-training is performed on a large-scale recognition database, and then fine-tuning training of model parameters is performed on the target facial expression recognition database, which alleviates the difficulty of model training due to the lack of large-scale data to a certain extent. Large differences in feature fields between the training and target datasets lead to unrobust features extracted by the model, and the recognition accuracy will also be negatively affected

[0007] Literature [2] (Y.Li, L.Song, X.Wu, R.He, and T.Tan. Antimakeup: Learning a bi-level adversarial network for makeup-invariant face verification. arXiv preprint arXiv: 1709.03654, 2017.) A method of data enhancement by generating face pictures with different poses and expressions is proposed, hoping to solve the problem of small data size and difficult training

However, Zhang et al. did not preserve the identity and expression information well when generating the target face image, resulting in a large degree of distortion in the generated image, and its authenticity and quality are relatively low

Images