Gesture recognition method and device based on large model, equipment and medium

By splicing the gesture image sequence into two-dimensional images and extracting features, combining the action rule library and time series model, the problem that traditional gesture recognition methods are difficult to recognize complex gesture action sequences is solved, and comprehensive semantic recognition of gesture image sequences is achieved.

CN119992643APending Publication Date: 2025-05-13ZHEJIANG ZHIPU XINPIAN TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411976544.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-12-31
Publication Date
2025-05-13

AI Technical Summary

Technical Problem

Traditional gesture recognition methods based on deep learning models are difficult to effectively identify the comprehensive semantics of complex gesture action sequences.

Method used

Using a large model-based gesture recognition method, by splicing the gesture image sequence into a two-dimensional image, global features and local features are extracted, and candidate rules are filtered from the action rule library using global features, local features and global rule features are fused, match scores between sequence features and candidate rules are calculated, and gesture recognition results are output.

Benefits of technology

It realizes effective recognition of complex gesture action sequences, can obtain the comprehensive semantics of gesture image sequences, and improves the accuracy of gesture recognition.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119992643A_ABST
    Figure CN119992643A_ABST
Patent Text Reader

Abstract

The invention relates to a gesture recognition method and device based on a large model, equipment and a medium, belongs to the technical field of gesture recognition, and solves the technical problem that comprehensive semantics of action sequences cannot be recognized in the prior art. The gesture recognition method based on the large model comprises the following steps: acquiring a to-be-recognized gesture image sequence; splicing the gesture image sequence into a two-dimensional image, and inputting the two-dimensional image into a gesture recognition model; extracting global features and local features of the two-dimensional image; according to the global features, candidate rules are screened from an action rule base, the global features and the candidate rules are fused to obtain global rule features, and the local features and the global rule features are fused to obtain sequence features corresponding to the candidate rules; and extracting context features of the sequence features, calculating matching degree scores of the context features and the corresponding candidate rules, and outputting a gesture recognition result according to the candidate rule with the highest matching degree score.
Need to check novelty before this filing date? Find Prior Art