A cerebellar neuronal network model applying to DIVA model

Abstract

The paper presents a new cerebellar feedback-error-learning model that meets the requirement of DIVA (Directions Into Velocities of Articulators) model, and constructs a cerebellar neuronal network model that consists of 900 granular cells, 10 Golgi cells, 50 basket/stellate cells, and 10 Purkinje cells based upon neuroanatomy. By applying the new cerebellar neuronal network model to the DIVA model and then modifying the cortical DIVA circuit, simulation results show that the DIVA model can produce clear and explicit speech sounds, and the new cerebellar neuronal network model can solve the problem of the learning and maintenance of feedforward motor commands.