Date of Award
1-1-2018
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Electrical Engineering
First Advisor
Reza Fazel-Rezai
Abstract
The goal of this research was to evaluate and compare three types of brain computer interface (BCI) systems, P300, steady state visually evoked potentials (SSVEP) and Hybrid as virtual spelling paradigms. Hybrid BCI is an innovative approach to combine the P300 and SSVEP. However, it is challenging to process the resulting hybrid signals to extract both information simultaneously and effectively. The major step executed toward the advancement to modern BCI system was to move the BCI techniques from traditional LED system to electronic LCD monitor. Such a transition allows not only to develop the graphics of interest but also to generate objects flickering at different frequencies. There were pilot experiments performed for designing and tuning the parameters of the spelling paradigms including peak detection for different range of frequencies of SSVEP BCI, placement of objects on LCD monitor, design of the spelling keyboard, and window time for the SSVEP peak detection processing. All the experiments were devised to evaluate the performance in terms of the spelling accuracy, region error, and adjacency error among all of the paradigms: P300, SSVEP and Hybrid. Due to the different nature of P300 and SSVEP, designing a hybrid P300-SSVEP signal processing scheme demands significant amount of research work in this area. Eventually, two critical questions in hybrid BCl are: (1) which signal processing strategy can best measure the user's intent and (2) what a suitable paradigm is to fuse these two techniques in a simple but effective way. In order to answer these questions, this project focused mainly on developing signal processing and classification technique for hybrid BCI. Hybrid BCI was implemented by extracting the specific information from brain signals, selecting optimum features which contain maximum discrimination information about the speller characters of our interest and by efficiently classifying the hybrid signals. The designed spellers were developed with the aim to improve quality of life of patients with disability by utilizing visually controlled BCI paradigms. The paradigms consist of electrodes to record electroencephalogram signal (EEG) during stimulation, a software to analyze the collected data, and a computing device where the subject’s EEG is the input to estimate the spelled character. Signal processing phase included preliminary tasks as preprocessing, feature extraction, and feature selection. Captured EEG data are usually a superposition of the signals of interest with other unwanted signals from muscles, and from non-biological artifacts. The accuracy of each trial and average accuracy for subjects were computed. Overall, the average accuracy of the P300 and SSVEP spelling paradigm was 84% and 68.5 %. P300 spelling paradigms have better accuracy than both the SSVEP and hybrid paradigm. Hybrid paradigm has the average accuracy of 79 %. However, hybrid system is faster in time and more soothing to look than other paradigms. This work is significant because it has great potential for improving the BCI research in design and application of clinically suitable speller paradigm.
Recommended Citation
Haider, Md. Ali, "Electroencephalogram Signal Processing For Hybrid Brain Computer Interface Systems" (2018). Theses and Dissertations. 2225.
https://commons.und.edu/theses/2225
Hybrid Experiment
P300.MOV (468300 kB)
P300 Experiment
SSVEP.MOV (218935 kB)
SSVEP Experiment