Title: A POMDP Approach to P300-Based Brain-Computer Interfaces
Authors: Jaeyoung Park, Kee-Eung Kim, and Sungho Jo
Venue: IUI 2010, Feb. 7-10 2010
Summary:
This paper discusses brain-computer interfaces (BCI) and computing an optimal sequence of flashesn in order to identify a desired selection using the fewest possible flashes. This is done through an EEG-based BCI. EEG is a non-invasive method that stands for encephalography. This BCI is perhaps the most popular method used. The system used is called a P300 speller which displays a 2x2 or 2x3 matrix to a user and then flashes the possible row. To help determine the best flash sequence is viewed as a decision making process and uses the POMDP model for this purpose.
The testing was done with the matrices describe above and showed to be more accurate over the baseline methods currently used in other BCI systems.
The image came from the paper itself.
Discussion:
This paper was interesting because it discussed a way to connect technology physically with the brain. This paper provides ways to improve the algorithms used that will allow faster, more accurate systems to be developed. The only problem I had with the paper was very technical discussion in the middle. It feels like unless you are an expert on the subject there is no way to follow what is being discussed. Other than that I thought the idea and improvements were fun to read about.
No comments:
Post a Comment