Pioneering research into Brain Computer Interfaces

Download 6.55 Mb.
Size6.55 Mb.
  1   2   3   4   5   6   7   8   9   ...   81

Pioneering research into

Brain Computer Interfaces

Mark Wessel

28 March 2006

Delft University of Technology

Faculty of Electrical Engineering, Mathematics, and Computer Science

Mediamatics: Man-Machine Interaction

Ir. Mark Wessel

Pioneering research into Brain Computer Interfaces
Delft University of Technology

Faculty of Electrical Engineering, Mathematics, and Computer Science

Mediamatics: Man-Machine Interaction

Mekelweg 4

2628 CD Delft

The Netherlands


Student number: 1015869

Keywords: brain computer interface, EEG, feature extraction
Copyright © 2005-2006 by Mark Wessel
A brain computer interface presents a direct communication channel from the brain. The BCI processes the brain activity and translates it into system commands using feature extraction and classification algorithms.

The overarching goal of this thesis is to make a start in the field of BCI. This thesis starts with an overview of the entire multi-disciplinary field of BCI, covering the basic fundamental ingredients, methods and modalities.

EEG-based BCI experiments have been designed and conducted. The experiments are designed to find distinctive brain patterns which are generated voluntary.

Next to the experiments an environment is created which offers a structured approach to the analysis of the EEG data from the experiments and allows for future continuation of BCI research.

The workbench contains, among others, the following models: ICA, FFT, AR, CSP and LD. The workbench performed well and produced quality results during testing. The quality of the experimental data after evaluation with the constructed workbench appeared to be mediocre at best, caused by the low spatial resolution of the EEG equipment, appliance errors and experimental design faults.

Recommendation for future work are to use different equipment, follow the line of synchronous BCI and construct a classifier to evaluate the data quality and construct an online BCI.

This thesis initiates BCI research at the TU Delft.

Abstract v

Contents vii

Definitions xii

Preface xiii

1.Introduction 1

1.1.Brain Computer Interface 1

1.2.The nature of EEG 1

1.3.Problem domain 2

1.4.Relevance 3

1.5.Scope 3

1.6.Research questions 3

1.7.Objectives 4

1.8.Outline 4

2.Theory 7

2.1.BCI overview 7

2.1.1.BCI definition 7

2.1.2.The history of BCI 7

2.1.3.The target group 7

2.2.Basic BCI elements 8

2.3.The input 9

2.3.1.The neuron 9

2.3.2.The brain 10

2.4. Brain activity measurement 11

2.4.1.EEG 13

2.4.2. Selecting a measurement method – Why EEG? 13

2.4.3. Invasiveness 13

2.4.4. The 10-20 system 14

2.4.5. Channels 14

2.5.The potential classes 14

2.5.1.Event related potentials 15

2.5.2.Rhythmic activity 15

2.5.3.Event related (de)synchronizations (ERD/ERS) 16

2.5.4. BCI approaches 16

2.5.5. Natural intent versus active control 16

2.5.6. Pattern recognition versus operant conditioning approach 17

2.5.7. Synchronous vs. asynchronous control 17

2.5.8. Universal vs. individual 18

2.5.9. Online vs. offline 18

2.6.Signal pre-processing 18

2.6.1.Amplification & A/D-converter 18

2.6.2.Filters 18

2.6.3. Reference filters 18

2.6.4. Bandpass filter 19

2.6.5.Artifacts 19

2.6.6.Artifact removal 20

2.6.7.Independent Component Analysis 20

2.6.8. Properties of ICA 21

2.6.9. Maximizing non-Gaussianity 21

2.6.10. The algorithm of ICA 22

2.6.11. Limitations of ICA 22

2.6.12.Purpose of ICA 22

2.6.13. ICA assumptions versus EEG 23

2.6.14.Channel selection 23

2.6.15.Segmentation 23

2.7.Translation algorithm 24

2.7.1.Feature extraction 24

2.7.2.Feature classification 24

2.8.Application 24

2.8.1.Training 25

2.8.2. Training principle 25

2.8.3. Feedback 26

2.9.Comparing BCI 27

2.9.1.BCI performance 27

2.9.2.Comparing criteria 27

2.9.3. Combining approaches 28

2.10.In depth look at the translation algorithm 28

2.10.1. Feature extraction 29

2.10.2. Fast Fourier Transform 29

2.10.5. Common Spatial Patterns 30

2.10.8.Parametric modelling 32

2.10.13.Discriminant analysis 34

2.10.19. Combining paradigms 36

2.10.20.Feature extraction conclusion 37

2.10.21. Feature classification 37

2.10.22. Genetic algorithm 37

2.10.25. Linear Vector Quantization 38

2.10.26. Neural Networks 39

2.10.27. Support Vector Machine 40

2.11.Overview of current BCI methods 41

2.12.Current software architecture 42

2.12.1. Software 42

2.12.2. Evaluation 43

2.13.Discussion and conclusion 44

3.Tools 46

3.1.Truscan EEG equipment 46

3.1.1.General description 46

3.1.2.System configuration 46

3.1.3.Usage 47

3.2.Truscan acquisition software 47

3.2.1.General description 47

3.2.2.Functionality 47

3.2.3.Remarks and conclusion 48

3.3.Truscan explorer software 48

3.3.1.General description 48

3.3.2.Functionality 49

3.3.3.Remarks and conclusion 50

3.4.IView X tool 50

3.4.1.System description 50

3.4.2.Configuration and usage 50

3.4.3.Remarks and conclusion 51

3.5.IView X software 51

3.5.1.General description 51

3.5.2.Configuration and usage 52

3.5.3.Remarks and conclusion 52

4.Experiment 54

4.1.Basic principle 54

4.2.Goals 54

4.3.Experimental design 54

4.3.1.Experimental design approach 54

4.3.2.Baseline task 55

4.3.3.Mental rotation 56

4.3.4.Motor imagery 56

4.3.5.Mathematical calculation 56

4.3.6.Visual presentation 57

4.3.7.Visual self selection 58

4.3.8.Visual and auditive presentation 58

4.3.9.Auditive presentation 58

4.3.10. Hyperventilation 59

4.4.Experiment setup 59

4.5.EEG cap setup 60

4.6.Artifacts 60

4.7.Task localization in brain 60

5.BCI Environment 62

5.1.BCI research environment requirements 62

5.2.Experiment tools 62

5.2.1.Experiment control program 62

5.2.2. Goal 62

5.2.3. System 62

5.2.4.Mouse control program 63

5.2.5.Goal 63

5.2.6.System 63

5.3.Data conversion 64

5.3.1.Conversion steps 64

5.3.2.Functional requirement 64

5.3.3. Non functional requirements 64

5.3.4.Design 64

5.3.5. Use case 65

5.3.6. Architecture 66

5.3.7.Implementation 67

5.3.8. Language 67

5.3.9. Function overview 67

5.4.Data storage design 67

6.E-Brain Design 70

6.1.Why a workbench? 70

6.2.System requirements 70

6.3.E-Brain functional requirements 71

6.3.1.Overview functional requirements - pre-processing 71

6.3.2.Overview functional requirements - analysis 71

6.3.3.Datastructure & data retrieval 71

6.3.4.Data down sampling 72

6.3.5.Basic pre-processing 72

6.3.6.Advanced pre-processing 74

6.3.7. Artifacts 74

6.3.8. Artifact rejection 74

6.3.9. Artifact removal 75

6.3.10.Data inspection 75

6.3.11.Data segmentation 75

6.3.12.Data analysis 75

6.3.13. FFT 76

6.3.14. AR 76

6.3.15. CSP 76

6.3.16. LD 77

6.3.17. Data management 77

6.4.Non functional requirements 77

6.5.Design 78

6.5.1.E-Brain use cases 78

6.5.2.Architecture 80

6.5.3.Dataflow 81

6.6.Usability 82

6.6.1.Mental model & metaphor 82

6.6.2.The interface 83

7.E-Brain implementation 84

7.1.Implementation environment 84

7.2. Detailed class diagram 84

7.3.Functionality 85

7.3.1.Datastructure & retrieval 85

7.3.2.Data inspection 87

7.3.3.Data down sampling 88

7.3.4.Basic pre-processing 89

7.3.5.Advanced pre-processing 90

7.3.6. Artifact removal 90

7.3.7. Finding the right ICA 92

7.3.8.Data segmentation 93

7.3.9.Data analysis 94

7.3.10.FFT 94

7.3.11. Frequency range 95

7.3.12. Data length 95

7.3.13. FFT 3D 96

7.3.14.AR 96

7.3.15. Model order estimation 97

7.3.16. Effects on the AR spectrum 98

7.3.17. AR approach 99

7.3.18. AR analysis 99

7.3.19. AR data averaging 100

7.3.20. Conclusion 101

7.3.21. CSP 102

7.3.22.CSP analysis 102

7.3.23.Interpreting CSP results 102

7.3.24.CSP settings 102

7.3.25. LD 103

7.3.26.LD analysis 103

7.3.27.LD settings 105

7.3.28. Data management 106

7.3.29. Interface & usability 106

8.E-Brain evaluation 110

8.1.Testing 110

8.1.1.Model functionality overview 110

8.1.2.Overview test sets 111

8.1.3.Test set 1 111

8.1.4.Results test set I for FFT & AR 112

8.1.5.Results test set I for CSP 113

8.1.6. Results test set I for LD 113

8.1.7.Test set 2 114

8.1.8.Test set 3 114

8.1.9. Results set 3 for CSP 115

8.1.10.Results set 3 for AR 117

8.1.11.Test set 4 119

8.1.12. Results set 4 for AR & FFT 119

8.1.13.Results set 4 for CSP 120

8.2.Workbench review 120

8.3.Conclusion 121

9.Results of experiments 122

9.1.Experiment result description 122

9.2.Experiment analysis 123

9.2.1.Pre-processing & analysis approach 123

9.2.2.Analysis subject one 124

9.2.3.Analysis subject two 129

9.3.Experiment evaluation 131

9.3.1.Discussion 131

9.3.2.Recommendations 132

9.3.3.Conclusion 133

10.Conclusion & recommendations 134

10.1.Conclusion & discussion 134

10.2.Recommendations & future work 135

Bibliography 138

Appendix A - Use cases 141

Appendix B – Planning 144

ALS Amyotrophic Lateral Sclerosis; a form of motor neuron disease

AR Autoregressive model

BCI Brain Computer Interface

CAR Common Average Reference

CSP Common Spatial Pattern

DFT Discrete Fourier Transform

E-Brain EEG-based BCI Research Analysis workbench

ECG Electrocardiography; measurement of electrical heart activity

ECoG Electrocorticography; invasive measurement of electrical brain activity

EEG Electroencephalography; measurement of electrical brain activity

EMG Electromyography; measurement of electrical muscle activity

EOG Electrooculargraphy; measurement of electrical ocular activity

ERP Event Related Potential

ERD Event Related Desynchronization

ERS Event Related Synchronization

ET Eye Tracker

FFT Fast Fourier Transform

FPE Akaike’s Final Prediction Error

GA Genetic Algorithm

GUI Graphical User Interface

IC Independent Component

ICA Independent Component Analysis

ITR Information Transfer Rate

LD Linear Discrimination

LVQ Linear Vector Quantization

MEG Magneto Encephalography; measurement of magnetic brain activity

MI Motor Imagery

MRI Magnetic Resonance Imaging; imaging using magnetic fields

MRP Movement Related Potential

NIR Near Infrared; imaging using near infrared light

NN Neural Network

PET Positron Emission Tomography; imaging using x-rays

ROC Receiver Operator Characteristic

SBC Schwarz’s Bayesian Criterion

SCP Slow Cortical Potential

SNR Signal to Noise Ratio

SQL Structured Query Language

SVM Support Vector Machine

VEP Visual Evoked Potential


Imagine sitting in a room and suddenly you would like to know every thing about the crusades in the Middle Ages, without moving a single muscle but by simply thinking about this idea triggers your brain interface system in searching the required data and transfers it directly to your brain. And you can start discussing it with your friends.

Imagine the old blind guy being able to see again by connecting a camera to his brain, although he lost his sight in a car accident years ago.

Imaging take a well deserved vacation to a tropical island by simply inserting the coordinates in your computer and your mind is off…

Many similar ideas have been uttered over the years and countless movies have been made concerning a link to and from the brain. Movies like the Matrix have always fascinated man. For years these stories belonged solely in the realm of science fiction; however the last couple of years a shift has been made in to the realm of reality.

While some proposed ideas may seem to be too futuristic to be possible, it is not unlikely that the principle idea behind it will become reality in the future. In fact it is already taking shape.

Science has always pursued the goal of aiding the human being in an ever increasing effort to increase the quality of life, whether this is to cure and aid the disabled, assist in high workload environment or simply for our pleasure and entertainment. It is apparent that the pursuit of technological advances and increasing science never stops.

Off course these techniques do not come falling from the sky, but require great effort from dedicated researchers all over the world. This is where we start our journey…

Chapter 1

  1. Introduction

    1. Brain Computer Interface

What is a Brain Computer Interface? As mentioned in the preface a BCI represents a direct interface between the brain and a computer or any other system. BCI is a broad concept and comprehends any communication between the brain and a machine in both directions: effectively opening a completely new communication channel without the use of any peripheral nervous system or muscles.

In principle this communication is thought to be two way. But present day BCI is mainly focusing on communication from the brain to the computer. To communicate in the other direction, inputting information in to the brain, more thorough knowledge is required concerning the functioning of the brain. Certain systems like implantable hearing-devices that convert sound waves to electrical signal which in turn directly stimulate the hearing organ already exist today. These are the first steps. The brain on the other hand is on a whole other complexity level compared to the workings of the inner ear.

From here on the focus is on communication directly from the brain to the computer. Most commonly the electrical activity (fields) generated by the neurons is measured, this measuring technique is known as EEG (Electroencephalography). An EEG-based BCI system measures specific features of the EEG-activity and uses these as control signals.

Over the past 15 years the field of BCI has seen a rapidly increasing development rate and obtained the interest of many research groups all over the world. Currently in BCI-research the main focus is on people with severe motor disabilities. This target group has little (other) means of communication and would be greatly assisted by a system that would allow control by merely thinking.

Directory: docs -> syllabi
docs -> Schulich School of Medicine & Dentistry Academic Promotion Teaching Dossier May 1, 2013 Dr. Test Doogie Howser
docs -> Vimala Mahmood Foundation
docs -> Dentist registration advice sheet Country of qualification: Romania
docs -> Minimum Requirements of Educational Programmes for the Acquisition of the Professional Qualification of Dentist, Pharmacist, Nurse and Midwife
docs -> Adopted by the Board of Registration in Dentistry, March 6, 2013; Amended June 5, 2013
docs -> Enhancing the Dental Public Health Workforce and Infrastructure Discussion Notes
docs -> Understanding Head Start Oral Health Program Information Report Data What is the Program Information Report (pir), and where can I find information about it?
docs -> Dental Services §17. 160 Authorization of dental examinations
syllabi -> Bayesian Networks Applied to Facial Expression Recognition August 2005, Paul Maaskant

Share with your friends:
  1   2   3   4   5   6   7   8   9   ...   81

The database is protected by copyright © 2019
send message

    Main page