A facial expression recognition method based on category difficulty evaluation and dynamic routing

By constructing a facial expression recognition model based on category difficulty assessment and dynamic routing, the model adaptively determines the difficulty of expression categories and performs differentiated processing for complex expressions. This solves the problem of ignoring category difficulty differences in existing technologies and improves recognition accuracy and efficiency.

CN122244918APending Publication Date: 2026-06-19NORTHWEST UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
NORTHWEST UNIV
Filing Date
2026-03-09
Publication Date
2026-06-19

AI Technical Summary

Technical Problem

Existing facial expression recognition methods ignore the differences in difficulty among expression categories, resulting in redundant calculations for simple expressions and decreased accuracy for complex expressions.

Method used

A facial expression recognition model based on category difficulty assessment and dynamic routing is constructed. Through semantic feature extraction, landmark feature extraction, window attention module, auxiliary classifier, category difficulty assessment unit, dynamic routing decision unit and cross-scale semantic enhancement unit, the model adaptively judges the difficulty of expression category and performs differentiated processing for complex expressions.

Benefits of technology

It improves the overall performance of facial expression recognition, especially the recognition accuracy of complex expressions, simplifies the processing of simple expressions, and achieves adaptive and efficient recognition.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122244918A_ABST
    Figure CN122244918A_ABST
Patent Text Reader

Abstract

This application relates to a facial expression recognition method based on category difficulty assessment and dynamic routing, comprising: constructing a facial expression recognition model, which includes a semantic feature extraction unit, a landmark feature extraction unit, a window attention module, an auxiliary classifier, a category difficulty assessment unit, a dynamic routing decision unit, a cross-scale semantic enhancement unit, and a classifier; training the facial expression recognition model based on a training dataset to obtain a trained facial expression recognition model; and inputting the facial expression image to be recognized into the trained facial expression recognition model to obtain the facial expression recognition result. This application, by introducing a category difficulty assessment mechanism, can adaptively determine the recognition difficulty of different expression categories, eliminating reliance on manual rules. Simultaneously, the dynamic routing design based on difficulty assessment can match differentiated processing paths for expressions of different difficulty levels, strengthen feature representation for complex expressions, and improve their recognition accuracy.
Need to check novelty before this filing date? Find Prior Art