Method, device and system for hand posture tracking

Abstract

The present invention provides a method, device and system for hand gesture tracking. The method includes acquiring a depth image including a hand region; according to the depth image, acquiring a target hand gesture through a target network model; wherein, the target The network model refers to a learning model that has been trained in advance and is used to perform feature extraction and feature recognition on the depth image to obtain the target initial hand pose, and optimize the target initial hand pose to output the target initial hand pose. Improve the accuracy and efficiency of attitude tracking. Real-time hand posture tracking can be realized without high computing device resources (such as GPU (Graphics Processing Unit, graphics processing unit)).

Description

technical field [0001] The present invention relates to the technical field of computer vision, in particular to a method, device and system for hand posture tracking. Background technique [0002] With the popularization of depth sensors and the demand in the field of human-computer interaction, hand gesture recognition and tracking based on depth data is one of the indispensable and important ways in new human-computer interaction technology, and has a wide range of application prospects, such as gesture remote control wireless Man-machine, gesture-controlled household robots, somatosensory games and assisted medical operations, etc. Compared with the traditional hand pose estimation based on RGB image, the depth data can provide the three-dimensional distance information of the hand. [0003] The existing hand gesture recognition in human-computer interaction requires the user's hand to be parallel to the imaging plane of the camera. However, people's hand gestures often...

AI Technical Summary

Problems solved by technology

[0003] The existing hand gesture recognition in human-computer interaction requires the user's hand to be parallel to the imaging plane of the camera. However, people's hand gestures often form a certain angle with the horizontal plane, which makes hand gesture recognition and tracking accuracy not high.

[0004] Or most of the existing hand tracking methods use fixed hand models. Different users have hands of different sizes. Using a fixed template hand will reduce the tracking accuracy. Sometimes it is necessary to pre-calibrate the hand shape according to different users, and the operation is cumbersome. , the attitude tracking efficiency is low, reducing the user experience

Images