Expression recognition method integrating global and local features of human face

Abstract

The invention discloses an expression recognition method integrating global and local features of a human face, and the method comprises the steps: designing three branches at a model feature extraction stage; enabling the first branch to use the whole face picture to extract the global features of the face; enabling the second branch to cut the face into two parts from top to bottom, enabling the third branch to cut the face into three parts from top to bottom, and extracting local features of the face; fusing the extracted global and local features, and inputting the fused features into a softmax classifier for face expression classification to obtain a recognition model integrating the global and local features of the face. Experiments show that the method of respectively extracting global and local features and then fusing the features can improve the accuracy of facial expression recognition.

Description

technical field [0001] The invention belongs to the fields of facial expression recognition, image processing and artificial intelligence, and in particular relates to an expression recognition method integrating global and local features of human faces. Background technique [0002] Facial Expression Recognition (FER) is an important means for machines to perceive human emotions and interact with humans, and it is an important research direction of human-computer interaction. Through the analysis of human facial expressions, it can roughly reflect the current emotions and potential intentions of people, which plays a decisive role in the social communication between people. At the same time, the advancement of facial expression recognition technology can promote the development of image processing, artificial intelligence and other fields. [0003] Traditional facial expression recognition usually first manually extracts facial features and then inputs them into a classifi...

AI Technical Summary

Problems solved by technology

[0004] In order to overcome the deficiencies of the prior art, the present invention provides an expression recognition method that integrates the global and local features of the human face. By dividing the human face into multiple parts, the global and local features of the human face are extracted to solve the problem of the prior art. Incomplete face feature extraction and over-fitting, gradient disappearance, and model degradation problems in deep network training further improve the accuracy of facial expression recognition

Images