视觉遮挡下的手势协调神经网络模型

A Gesture Coordinated Neural Network Model under Visual Occlusion

在人类抓握运动过程中,相比于目标物体没被遮挡,目标物体被遮挡时进行抓取物体会导致更大的抓握孔径。然而,以前没有用于解释这种效应机制的抓握运动神经网络模型。针对这种情况,对视觉未被遮挡时进行抓握的Vilaplana模型进行研究与分析,同时讨论了视觉遮挡这种特殊情况对手臂移动与抓取的影响。在此基础上,将手臂移动与手势抓取划分为四个阶段,即手臂收缩阶段、抓取物体阶段,手臂关节收缩阶段、物体释放阶段。讨论视觉对各个阶段的影响程度,对Vilaplana的模型进行改进,使之能够适应视觉遮挡下的抓取运动,最终提出了视觉遮挡下的手势协调模型。该模型通过增大峰值抓握孔径来补偿视觉的不确定性,符合一般的人体机理。通过MATLAB仿真证明,视觉遮挡并不会对新模型的手臂运动产生较大的影响,抓握孔径的增大将会避免手与物体之间不必要的碰撞。

In the process of human grasping,compared to the object that is not obscured,the capture of the object when it is blocked will lead to a greater grasp of the grip aperture.However,there has not been a grip neural network model that has been used to explain this mechanism before.In view of this situation,we study and analyze the Vilaplana model of gripping without visual occlusion.At the same time,we discuss the influence of visual occlusion on arm movement and grasping.On this basis,arm movement and gesture grasping are divided into four stages,namely,arm contraction stage,object grasping stage,arm joint contraction stage and object releasing stage.The influence of vision on each stage is discussed.Vilaplana model is improved to adapt to the grasping motion under visual occlusion.Finally,a gesture coordination model under visual occlusion is proposed.The model compensates for the uncertainty of the vision by increasing the aperture of the peak grip,which is in line with the general human body mechanism.Through MATLAB simulation,it is proved that visual occlusion will not have a great impact on the arm movement of the new model,and the increase of grasping aperture will avoid unnecessary collision between hands and objects.