Facial expression recognition method combining LBP features and lightweight neural network

Abstract

The invention provides a facial expression recognition method combining LBP features and a lightweight neural network. In the recognition method, a full convolution layer is split into two sub-convolution of deep convolution and point-by-point convolution by adopting an idea of Xception model decomposition to extract features, so that the number of parameters and the calculation cost are greatly reduced; meanwhile, in order to solve the problems of gradient disappearance and gradient explosion and improve the propagation capacity of the gradient between product layers, an inverted residual structure with a linear bottleneck in MobileNetV2 is adopted in the method; moreover, due to the fact that the full-connection layer is prone to overfitting, severe dependence on dropout regularization is caused, and the generalization ability of the whole network is affected, the CNN full-connection layer is replaced with global average pooling, that is, all pixel values of the feature map of each channel are averaged, and then the obtained feature vectors are directly sent to a Softmax layer for classification, so that not only is the correspondence between feature mapping and categories enhanced, but also the over-fitting of the overall structure is prevented.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a facial expression recognition method combining LBP features and a lightweight neural network. Background technique [0002] With the change of interaction methods, the research on facial expression recognition is attracting more and more attention. According to research, the information conveyed by facial expressions accounts for a very large proportion, as high as 55%, of which 38% comes from the speaker's tone, voice, rhythm, etc., and only 7% depends on the content of the speaker. As a form of non-verbal communication, facial expressions can convey non-verbal information, which can be used as language aids to help listeners infer the speaker's emotional changes and intentions. It can be seen that facial expressions play a vital role in human-to-human communication. In 1971, psychologist Ekman and others first proposed that human beings have six main emotions, and ...

AI Technical Summary

Problems solved by technology

[0005] In view of this, the purpose of the present invention is to provide a facial expression recognition method that combines LBP features and a lightweight neural network, aiming to solve the problems of poor robustness and weak generalization ability of the recognition model that exists in the prior art , improve the accuracy of expression recognition in a scene that is more in line with real life

Images