Neural Network Expression Recognition Method Based on Graph Structure

Abstract

The present invention relates to the field of biological feature recognition, specifically a neural network expression recognition method based on a graph structure, comprising: locating multiple key points for facial expression recognition; using a filter to extract the texture feature vector of each key point; Texture feature vectors are used as nodes, and each node is connected to form a graph structure, and the connection lines between each key point are used as edges of the graph structure, and the geometric feature information between each key point is represented by Euclidean distance, and the graph structure It is used to replace the expression image; the graph structure is input into the BRNN neural network; and the result of expression classification is output on the last time step of the BRNN neural network by using multiple iterations. The invention makes full use of the texture and geometric feature information of different expressions to solve the problem that the accuracy of expression recognition is reduced due to the interference information on the image.

Description

technical field [0001] The invention relates to the field of biological feature recognition, in particular to a neural network expression recognition method based on a graph structure. Background technique [0002] In the research of facial expression recognition, extracting effective expression information from face images is a key step. Early expression recognition research mainly starts from three aspects: feature learning, feature extraction and separator construction. First, people extract information about facial appearance or geometry changes from image or video series, represented by methods such as LBP-TOP, HOG3D, DTAGN and STM-ExpLet. Then, a subset of features that can effectively represent facial expressions are selected from them, and finally, according to the extracted features, an effective classifier is constructed to recognize facial expressions. However, relatively speaking, traditional classification learning methods are difficult to achieve better result...

AI Technical Summary

Problems solved by technology

[0005] The present invention intends to provide a neural network facial expression recognition method based on a graph structure, making full use of the texture and geometric feature information of different facial expressions, so as to solve the problem that the accuracy of facial expression recognition is reduced due to the interference information existing on the image

Images