Multi-Grained Gesture Modeling for Emotion Recognition

Read full text Summarise with VIRAT
Authors: Rong Wei, Xuepeng Huang
Cited by: 0
References: 52

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

In recent years, emotion recognition from body gesture movements has become an important research field with the continuous development and improvement of depth sensors and computer vision algorithms. Most of the public datasets are records of continuous emotional actions, and frame by frame analysis or as a single gesture recognition will ignore the context relationship between actions. Thus, the method based on manual feature extraction performs poorly on large-scale datasets. To solve the above problems, this paper proposes an effective gesture emotion recognition model. The model is composed of pseudo energy model and multi-grained model. The pseudo energy model quantifies the energy of gesture and provides an important basis for key frame extraction. The multi-grained model is proposed on the basis of both fine grained which provides specific information and coarse grained which provides logical information. By integrating multiple levels of granularity into a unified architecture, we demonstrate enhanced classification performance. Our method surpasses the prior state-of-the-art approaches in two datasets and achieves the highest 82.7% classification accuracy in the KDAE dataset and 80.6% classification accuracy in the EMBG dataset. The excellent experimental results on two public datasets show that it is feasible to divide the emotion of action from the perspective of energy and use multi-grained model to provide the low-level and high-level features required by the network. In the task of gesture emotion recognition, our method is obviously superior to the state-of-the-art methods.

Fields of Study & Subjects

Emotion and Mood Recognition Hand Gesture Recognition Systems Human Pose and Action Recognition

Publication details

Journal / source
IEEE Access
Volume
14
Pages
67046–67059
Publication date
1 January 2026
Language
EN