Development of the 'Mirror System': A Computational Model

摘要

The authors are studying the development of the mirror system from a computational perspective with the ultimate goal of realizing a physical implementation. They are using an anthropomorphic robot as a development platform. The body of the robot provides the physical interaction between the computational structure and the environment. Single neuron recording and microstimulation in the monkey have shown that the premotor cortex contains visually responsive neurons. These neurons are thought to encode visual characteristics of objects in motor terms. For example, area F5 contains neurons that respond both when the monkey grasps an object and when it only fixates on the object (canonical neurons). Regardless of the intention of actually performing that particular motor act, the brain keeps a representation of the potential motor act. F5 also contains another class of visuomotor responsive cells called 'mirror neurons.' They too fire when manipulating an object by also when one is watching someone else performing the same manipulation. However, pretending to do the gesture does not make the neurons fire; the action has to be goal-directed. F5 can be imagined as the result of a two-stage process. The paper also lists the visual competencies and motor behaviors that have already been implemented.