Fine motion classification and regression
By introducing space-time video datasets and machine learning models, the limitations of the existing technology in fine motion evaluation are solved, and the fine classification and regression of human actions are achieved, and the accuracy and granularity of action evaluation are improved.
Patent Information
- Application Number
- CN202480002727.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2024-10-22
- Filing Date
- 2024-10-29
- Publication Date
- 2025-03-11
AI Technical Summary
Existing video action recognition technologies have limitations when more sophisticated action evaluation is required, making it difficult to accurately classify or rate actions under specific program standards.
By introducing space-time video datasets, using processors and nonvolatile computer-readable storage media, a method of fine action classification and regression is realized. The method includes receiving a video stream, identifying a reference object, extracting poses and object data sets, generating composite data structures, and inputting them into a trained machine learning model for classification and regression.
The fine classification and regression of human actions is achieved, and the actions can be accurately evaluated according to specific procedural standards, which improves the granularity and accuracy of action evaluation.
Smart Images
Figure CN119604907A_ABST
Abstract
Citation Information
Patent Citations
System and method for human gait analysis
US20210275107A1
Deep-learning-based fall detection based on human keypoints
US20220079472A1
Image analysis using neural networks for pose and action identification
US20240037977A1