To address the limitations of existing gesture-based human-robot interaction (HRI) methods

which focus on static sign language and pose tracking and cannot support the control of complex robotic systems

a remote interaction system for quadruped robots that integrates an RGB-D camera sensor and dynamic gestures is proposed.Data are acquired using an RGB-D camera sensor

and a joint inference framework combining RGB and depth images is employed to accurately extract the 3D coordinates of 21 hand skeletal key points.A semantic-pose to motion (SPM) model is constructed

and the semantic and pose information of the dynamic gestures is analyzed simultaneously.The 3D coordinates are then converted into robot executing actions in real time

enabling integrated control of quadrupedal movement

manipulator end tracking

and command switching.Real-environment experiments demonstrate that the system can achieve stable real-time interaction and precise control.This method effectively overcomes the shortcomings of traditional static gesture interaction

providing a natural and efficient non-contact interaction solution for complex robots

and significantly improving the usability and operational flexibility of complex robot platforms.