Acquiring Bidirectional Computation Model of Motion Recognition and Generation and Its Application to Humanoids

Abstract

The discovery of Mirror Neurons in human brain shows that the motion recognition and generation are executed by bidirectional computation model, and the model leads to symbol grounding system through embodied humanoids. We have proposed mirror neuron models based on Hidden Markov Models. In this paper, we propose a new method for usual models in order to be developed through behavior observation and generation.