Non-Linear Bayesian System Identification of Cortical Responses Using Volterra Series

Abstract

The human sensorimotor system can be seen as a complex network in which the brain plays an important role, resulting in a difficult-to-understand relation between proprioceptive stimuli and cortical responses. However, understanding this relationship is of added value for understanding various diseases which cause dysfunctionality. In recent years, a variety of studies have been conducted towards finding the non-linear relationship between the cortical responses and wrist joint manipulation. This research is dedicated to providing an initial set-up to create models that are able to provide accurate predictions despite noisy data. The relationship between wrist joint manipulation and the cortical response is assumed to be non-linear and the corresponding identification method is categorized in a two-step process, namely the model structure, i.e. Volterra series, and stochastic identification method, i.e. Bayesian Inference. To understand the working principle of the proposed algorithm, the method is first applied to a set of computer models. Finally, an attempt is made to model the cortical responses evoked by wrist joint manipulations