Date of Submission

Spring 2012

Academic Program

Computer Science

Project Advisor 1

Sven Anderson

Abstract/Artist's Statement

A Brain Computer Interface (BCI) is a computer-based system that attempts to translate thoughts into computational commands. Thought is represented by the brain-activity of the user, which is recorded using electroencephalography (EEG) or magnetoencephalography (MEG). The system learns to identify input patterns and to associate them with computer commands.

In this study, a BCI is designed that includes a pattern recognizer (PR) for feature extraction, and a hill-climbing optimizer for translating signals into commands.

Evaluation of the pattern recognizer demonstrates the ability to correctly recognize patterns that vary by in spatial scale or with up to noise. Both creation and identification of patterns is accurate for input data having no more than noise, degrading to approximately accuracy for noise. Increasing the threshold of spatial scale tolerance resulted in approximately the same net error, but with an increase in the ratio of type 1 errors to type 2 errors. Similarly, decreasing the threshold of spatial scale tolerance resulted in a decrease in the ratio of type 1 errors to type 2 errors. A decrease in error was elicited when the system was given enough memory and data that the signal processor could compensate for noise by generating large numbers of memories.

Distribution Options

Access restricted to On-Campus only

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 License.