Hand action recognition method, device and equipment based on 3D key point sequence

Through the hand movement recognition method based on 3D key point sequence, the problem of high cost of obtaining depth information and inaccurate recognition results in the prior art is solved, and the accurate recognition of multiple people and multiple hand movements is achieved, reducing the cost and improving the recognition accuracy.

CN120071439APending Publication Date: 2025-05-30BEIJING YUANLUO TECHNOLOGY CO LTD
0 Cites 0 Cited by

Patent Information

Application Number
CN202510135994.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-02-07
Publication Date
2025-05-30

AI Technical Summary

Technical Problem

The existing hand motion recognition technology needs to rely on wearable devices or motion capture when obtaining depth information. The stepping cost is high and there are limitations on application scenarios. The fixed intercept of the action frames leads to inaccurate recognition results, and it is impossible to accurately identify actions with insignificant local changes.

Method used

The hand action recognition method based on the 3D key point sequence is adopted, and the 3D key point coordinates of the hand are determined by obtaining the depth information of the video stream and performing hand detection, and determining whether the hand action is over based on the movement speed, thereby generating an adaptive length action sequence and inputting it into the pre-trained hand action recognition model for identification.

Benefits of technology

The recognition of multiple people and multiple hand movements is realized, the annotation and hardware costs are reduced, and the accuracy of action recognition is improved. Especially in the motion recognition where local changes are not obvious, it makes up for the defects of the existing 2D key point detection model.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention provides a hand action recognition method, device and equipment based on a 3D key point sequence, and relates to the technical field of action recognition, and the method comprises the steps: obtaining the depth information of each video frame, and carrying out the hand detection of the video frames, and obtaining a plurality of detection frames; determining hand 3D key point coordinates of the ID corresponding to the detection frame under the video frame based on the detection frame image and the depth information; for any ID, determining the hand movement speed according to the hand 3D key point coordinates of the ID under each video frame; if the hand movement speed meets the configured termination rule, generating a hand 3D key point sequence of the ID based on the hand 3D key point coordinates of the ID under the plurality of continuous video frames; and inputting the hand 3D key point sequence into a pre-trained hand action recognition model to obtain a hand action recognition result corresponding to the ID. According to the invention, the method can achieve the recognition of a plurality of hand actions of a plurality of people in a long-range scene and / or a close-range scene.
Need to check novelty before this filing date? Find Prior Art