A dynamic gesture recognition method, device and equipment
By acquiring hand image sequences for region detection and optical flow analysis, combined with three-dimensional residual network processing, the problem of poor real-time performance of existing dynamic gesture recognition methods is solved, and efficient dynamic gesture recognition is achieved.
CN116798110BActive Publication Date: 2026-05-29CHINA MOBILE COMM LTD RES INST +2
Patent Information
- Authority / Receiving Office
- CN Β· China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA MOBILE COMM LTD RES INST
- Filing Date
- 2022-03-14
- Publication Date
- 2026-05-29
AI Technical Summary
Technical Problem
Existing dynamic gesture recognition methods rely on two-dimensional convolutional neural networks or deep image processing, resulting in poor real-time recognition performance and difficulty in recognizing complex movements.
Method used
By acquiring hand image sequences, hand region detection and optical flow analysis are performed, and combined with a three-dimensional residual network for fusion image processing to recognize dynamic gestures.
Benefits of technology
It achieves strong generalization ability, strong expressive ability, and good scalability, and uses as few parameters as possible while ensuring accuracy, thus improving recognition speed and classification effect.
β¦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN116798110B_ABST
Abstract
The application provides a dynamic gesture recognition method, device and equipment. The method comprises the following steps: acquiring a hand image sequence; the hand image sequence comprises N frames of to-be-detected hand images; wherein N>1, and N is an integer; performing hand region detection on the to-be-detected hand images to obtain N frames of detection images corresponding to the to-be-detected hand images; performing optical flow analysis processing on adjacent two frames of to-be-detected hand images to obtain N-1 frames of optical flow grayscale images; performing fusion image processing on the to-be-detected hand images, the detection images and the optical flow grayscale images to obtain a fusion image sequence; inputting the fusion image sequence into a three-dimensional residual network to obtain a dynamic gesture classification and a dynamic gesture category recognition result; the scheme of the application has strong generalization ability, strong expression ability and good expansibility; in the case of ensuring the accuracy, the model uses as few parameters as possible, ensures the recognition speed of the model, and has better classification effect.
Need to check novelty before this filing date? Find Prior Art