Human posture recognition methods, devices, computer equipment, and storage media

By constructing a human pose recognition model based on a deep convolutional neural network, the problem of limited computing and storage resources was solved, and the accuracy and efficiency of human pose recognition in complex backgrounds were improved.

CN118334744BActive Publication Date: 2026-05-26SOUTH CHINA NORMAL UNIV
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202410496744.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-04-24
Publication Date
2026-05-26
Estimated Expiration
2044-04-24

AI Technical Summary

Technical Problem

Existing human pose recognition models in complex backgrounds are limited by computational and storage resources, making it difficult to perform effective and accurate human pose recognition, especially when there is occlusion or changes in lighting conditions, which increases the difficulty of recognition.

Method used

By constructing a human pose recognition model, iterative training is performed using a deep convolutional neural network, including a downsampling network, a backbone network, an attention extraction network, a deep feature extraction network, and a joint matching network. Deep convolutional feature maps are extracted, joint matching and training are performed, and information about the relevant regions for human pose estimation is adaptively captured.

Benefits of technology

Despite limited computing and storage resources, this technology improves the accuracy and performance of human pose recognition, enabling effective human pose recognition in complex contexts.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118334744B_ABST
    Figure CN118334744B_ABST
Patent Text Reader

Abstract

This invention relates to the field of human pose recognition, and particularly to a human pose recognition method, apparatus, computer device, and storage medium. By extracting deep convolutional feature maps from several sample human pose images as training data, a human pose recognition model is iteratively trained to construct a target human pose recognition model. While ensuring the accuracy of human pose recognition, the model can adaptively capture the importance of information related to human pose estimation regions, further improving the model's performance. This enables the model to perform effective and accurate human pose recognition even with limited computing and storage resources.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Method and apparatus for estimating pose of human body

    CN113892113A

  • Action point-based behavior recognition method, electronic equipment and storage medium

    CN115147935A

  • Human body posture estimation model training method, human body posture estimation method, human body posture estimation device and electronic equipment

    CN117037215A