A stochastic control approach to optimally designing hierarchical flash sets in P300 communication prostheses

Rui Ma, Navid Aghasadeghi, Julian Jarzebowski, Timothy Bretl, Todd P. Coleman

Research output: Contribution to journalArticle

Abstract

The P300-based speller is a well-established brain-computer interface for communication. It displays a matrix of objects on the computer screen, flashes each object in sequence, and looks for a P300 response induced by flashing the desired object. Most existing P300 spellers uses a fixed set of flash objects. We demonstrate that performance can be significantly improved by sequential selections from a hierarchy of flash sets containing variable number of objects. Theoretically, the optimal hierarchy of flash sets - with respect to a given statistical language model - can be found by solving a stochastic control problem of low computational complexity. Experimentally, statistical analysis demonstrates that the average time per output character at 85% accuracy is reduced by over 50% using our variable-flash-set approach as compared to traditional fixed-flash-set spellers.

Original languageEnglish (US)
Article number6112238
Pages (from-to)102-112
Number of pages11
JournalIEEE Transactions on Neural Systems and Rehabilitation Engineering
Volume20
Issue number1
DOIs
StatePublished - Jan 1 2012

Keywords

  • Brain-computer interfaces (BCIs)
  • P300
  • hierarchical flash sets
  • stochastic dynamic programming

ASJC Scopus subject areas

  • Internal Medicine
  • Neuroscience(all)
  • Biomedical Engineering

Fingerprint Dive into the research topics of 'A stochastic control approach to optimally designing hierarchical flash sets in P300 communication prostheses'. Together they form a unique fingerprint.

  • Cite this