Date of Award

Fall 2009

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Biomedical Engineering

First Advisor

Steven A. Jones

Abstract

For persons with severe disabilities, a brain computer interface (BCI) may be a viable means of communication, with scalp-recorded electroencephalogram (EEG) being the most common signal employed in the operation of a BCI. Various electrode configurations can be used for EEG recording, one of which was a set of concentric rings that was referred to as a Laplacian electrode. It has been shown that Lapalacian EEG could improve classification in EEG recognition, but the complete advantages of this configuration have not been established.

This project included two parts. First, a modeling study was performed using Independent Component Analysis (ICA) to prove that tripolar electrodes could provide better EEG signal for BCI. Next, human experiments were performed to study the application of tripolar electrodes in a BCI model to show that the application of tripolar electrodes and data-segment related parameter selection can improve EEG classification ratio for BCI.

In the first part of work, an improved four-layer anisotropic concentric spherical head computer model was programmed, then four configurations of time-varying dipole signals were used to generate the scalp surface signals that would be obtained with tripolar and disc electrodes. Four important EEG artifacts were tested: eye blinking, cheek movements, jaw movements and talking. Finally, a fast fixed-point algorithm was used for signal-independent component analysis (ICA). The results showed that signals from tripolar electrodes generated better ICA separation than signals from disc electrodes for EEG signals, suggesting that tripolar electrodes could provide better EEG signal for BCI.

The human experiments were divided into three parts: improvement of the data acquirement system by application of tripolar concentric electrodes and related circuit; development of pre-feature selection algorithm to improve BCI EEG signal classification; and an autoregressive (AR) model and Mahalanobis distance-based linear classifier for BCI classification. In the work, tripolar electrodes and corresponding data acquisition system were developed. Two sets of left/right hand motor imagery EEG signals were acquired. Then the effectiveness of signals from tripolar concentric electrodes and disc electrodes were compared for use as a BCI. The pre-feature selection methods were developed and applied to four data segment-related parameters: the length of the data segment in each trial (LDS), its starting position (SPD), the number of trials (NT) and the AR model order (AR Order). The study showed that, compared to the classification ratio (CR) without parameter selection, the CR was significantly different with an increase by 20% to 30% with proper selection of these data-segment-related parameter values and that the optimum parameter values were subject-dependent, which suggests that the data-segment-related parameters should be individualized when building models for BCI. The experiments also showed that that tripolar concentric electrodes generated significantly higher classification accuracy than disc electrodes.

Download

Share

COinS