A stochastic control framework for the design of observational brain-computer interfaces based on human error potentials

Brain-computer interfaces (BCI) allow human subjects to interact with exogenous systems through neural signals. The goal of this interaction may be to induce behaviors or properties in either the exogenous system or the human subject. In this thesis, we develop a novel framework for designing BCIs based on principles from adaptive control. In particular, we exploit scalp electroencephalography-derived correlates of the human error processing system to recover a subject’s desired policy for the exogenous system. This scheme allows a human subject to control a system through passive observation by critiquing actions taken by the system. We provide a necessary and sufficient condition for convergence and simulations as a proof of concept. Further, we discuss the application of this framework to building co-adaptive BCIs and as a tool for understanding the learning process during BCI interaction.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Electrical & Computer Engr

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2011

Author dc:creator

• Steines, David A.

Contributors dc:contributor

• Coleman, Todd P.

• Brain-computer interface
• feedback-related negativity
• Bayesian control rule

Statement dc:rights

• Copyright 2011 David A. Steines

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/24378

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/24378