Fast and Accurate Hand Visual Detection by Using a Spatial-Channel Attention SSD for Hand-Based Space Robot Teleoperation

Table 1

Comparison and summary of the hand-based robot teleoperation methods.

Method	Brief methodology	Highlights	Limitations
Haptic feedback-based method [3–5]	This method transmits the 6-Dof position and orientation of human hand to the robot through the haptic feedback controller.	High accuracy and mature technology.	Contact method and only be used for 6-DOF control.
sEMG-based method [10–12]	When hand moves, the arm will generate corresponding motor neuron information, which can be obtained by decoding the sEMG signal.	High accuracy.	Contact method and lack of robustness.
Data glove-based method [7–9]	This method uses curvature sensors to collect the bending degrees of the fingers and the posture change of the entire human hand, to decode the movement of the hand.	High accuracy and mature technology.	Contact method and lack of robustness.
Vision-based method [14–16]	This method uses camera to capture the movements of human hands and maps the human hand movement information to the robot.	Strong noise and immature technology.	Noncontact method, strong robustness, and naturalness.