Bayesian Classification with Split-and-merge Gaussian Process (SMGP) Prior in EEG-based Brain-Computer Interfaces Restricted; Files Only

Abstract

An Event-Related Potential (ERP)-based Brain-Computer Interface (BCI) Speller System is a device that interprets brain activity to operate technology. The system works by searching target (relevant) signals among a set of non-target (irrelevant) ones via electroencephalogram (EEG). We developed a Bayesian generative framework to model EEG signals at the stimulus level. We performed feature selection based on a Split-and-merge Gaussian Process (SMGP) prior to identify the spatial-temporal differences between target and non-target ERP responses. This greatly improved our ability to accurately detect P300 ERPs. Our findings from both simulation studies and real data analysis demonstrated that stimulus-level analysis enhanced P300 detection and revealed neural dynamics, which could inform the development of BCIs with advanced predictive and personalized capabilities.

Table of Contents

1 Introduction 1

2 Bayesian Modeling of EEG-BCI Data 4

2.1 Problem Setup and Notations 4

2.2 A Bayesian Model 5

2.3 The Split-and-Merge GP 5

3 Posterior Inference 7

3.1 Model Representation and Prior Specification 7

3.2 Markov Chain Monte Carlo 9

3.3 Posterior Character-Level Probability for Prediction 10

4 Simulation Studies 11

4.1 Setup 11

4.2 Model Fitting and Diagnostics 11

4.3 Results 12

4.3.1 Parameter Estimation 12

4.3.2 Prediction Performance 14

5 Analysis of Real EEG-BCI Data 15

5.1 Dataset and Pre-processing 15

5.2 Model Setting 16

5.3 Results 17

5.3.1 Parameter Estimation 17

5.3.2 Prediction Performance 19

6 Discussion 21

7 References 24

About this Master's Thesis

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Rollins School of Public Health
Department	Biostatistics
Subfield / Discipline	Biostatistics - MPH & MSPH
Degree	M.S.P.H.
Submission	Master's Thesis
Language	English
Research Field	Biology, Biostatistics Statistics
Parola chiave	Brain-Computer Interface (BCI) Bayesian analysis Gaussian process
Committee Chair / Thesis Advisor	Tianwen Ma, Emory University
Committee Members	Joshua Lukemire, Emory University

Ultima modifica

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	File download under embargo until 15 May 2027	2025-04-17 00:52:39 -0400	File download under embargo until 15 May 2027

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions
	File download under embargo until 15 May 2027	2025-04-14 18:46:01 -0400	File download under embargo until 15 May 2027