【摘要】 假设人脑存储着肢体的几何和动力学模型 ,以一个多层感知网络代表这个内部模型 ,仿真了在三种不同情况下人体肘关节韵律运动 :1)内部模型的参数与肢体的几何和动力学模型参数完全匹配 ,2 )上述参数不匹配 ,3 )存在外界干扰 .仿真结果显示模型的参数不匹配是导致肢体韵律运动误差的主要原因 .由此可见 ,运动学习过程就是内部模型不断更新以匹配肢体几何和动力学模型的过程更多还原

【Abstract】 Human brain is hypothesized to store a geometry and dynamic model of the limb.A multilayer perceptron (or MLP) network is used to stand for the model.In this paper the human elbow joint rhythmic movement is simulated in three cases:1)Parameters of the MLP,the limb geometry and dynamic model match completely,2)Parameters mismatch between them,and 3)Disturbance exists.The results show that parameters mismatch is the main error source,which causes the elbow joint movement to be aberrant.From this we can infer that movement study is a process in which the internal model is updated continuously to match the geometry and dynamic model of limb.更多还原