Technological University Dublin ARROW@TU Dublin Dissertations School of Computer Sciences 2019 Noise Reduction in EEG Signals using Convolutional Autoencoding Techniques Conor Hanrahan Technological University Dublin Follow this and additional works at: https://arrow.tudublin.ie/scschcomdis Part of the Computer Engineering Commons, and the Computer Sciences Commons Recommended Citation Hanrahan, C. (2019) Noise Reduction in EEG Signals using Convolutional Autoencoding Techniques, Masters Thesis, Technological University Dublin. This Theses, Masters is brought to you for free and open access by the School of Computer Sciences at ARROW@TU Dublin. It has been accepted for inclusion in Dissertations by an authorized administrator of ARROW@TU Dublin. For more information, please contact [email protected], [email protected]. This work is licensed under a Creative Commons Attribution-Noncommercial-Share Alike 4.0 License Noise Reduction in EEG Signals using Convolutional Autoencoding Techniques Conor Hanrahan A dissertation submitted in partial fulfilment of the requirements of Technological University Dublin for the degree of M.Sc. in Computer Science (Data Analytics) September 2019 I certify that this dissertation which I now submit for examination for the award of MSc in Computing (Data Analytics), is entirely my own work and has not been taken from the work of others save and to the extent that such work has been cited and acknowledged within the test of my work. This dissertation was prepared according to the regulations for postgraduate study of the Technological University Dublin and has not been submitted in whole or part for an award in any other Institute or University. The work reported on in this dissertation conforms to the principles and requirements of the Institute’s guidelines for ethics in research. Signed: _________________________________ Date: 01 September 2019 i ABSTRACT The presence of noise in electroencephalography (EEG) signals can significantly reduce the accuracy of the analysis of the signal. This study assesses to what extent stacked autoencoders designed using one-dimensional convolutional neural network layers can reduce noise in EEG signals. The EEG signals, obtained from 81 people, were processed by a two-layer one-dimensional convolutional autoencoder (CAE), whom performed 3 independent button pressing tasks. The signal-to-noise ratios (SNRs) of the signals before and after processing were calculated and the distributions of the SNRs were compared. The performance of the model was compared to noise reduction performance of Principal Component Analysis, with 95% explained variance, by comparing the Harrell-Davis decile differences between the SNR distributions of both methods and the raw signal SNR distribution for each task. It was found that the CAE outperformed PCA for the full dataset across all three tasks, however the CAE did not outperform PCA for the person specific datasets in any of the three tasks. The results indicate that CAEs can perform better than PCA for noise reduction in EEG signals, but performance of the model may be training size dependent. Key words: electroencephalography, event-related potential, noise in EEG, signal-to- noise ratio, noise reduction, artifact removal, convolutional autoencoder ii ACKNOWLEDGEMENTS I would like to express my sincere thanks to Dr. Luca Longo for his assistance and guidance throughout the course of this dissertation. I would like to thank Alexander Suvorov who, through a series of discussion, aided me with numerous aspects of this undertaking, which allowed me to perform to the best of my ability throughout this process. I would also like to acknowledge and convey my thanks to the rest of the Technological University Dublin staff that have assisted and supported me throughout my time in the university. Finally, without the support and encouragement of my family I would not have been able to complete this undertaking. I’m eternally grateful. iii TABLE OF CONTENTS ABSTRACT ................................................................................................................ II TABLE OF CONTENTS .......................................................................................... IV TABLE OF FIGURES ............................................................................................ VII TABLE OF TABLES ................................................................................................ IX 1 INTRODUCTION ............................................................................................... 10 1.1 BACKGROUND .................................................................................................. 10 1.2 RESEARCH PROJECT ......................................................................................... 11 1.3 RESEARCH OBJECTIVES .................................................................................... 12 1.4 RESEARCH METHODOLOGIES ........................................................................... 12 1.5 SCOPE AND LIMITATIONS ................................................................................. 14 1.6 DOCUMENT OUTLINE ....................................................................................... 15 2 REVIEW OF EXISITING LITERATURE ...................................................... 16 2.1 BRAIN ACTIVITY .............................................................................................. 16 2.1.1 Event Related Potentials ....................................................................... 16 2.2 EEG ................................................................................................................. 18 2.3 NOISE IN EEG .................................................................................................. 22 2.3.1 Definition & Background ..................................................................... 22 2.3.2 Causes of noise ..................................................................................... 22 2.3.3 Signal-to-noise ratio ............................................................................. 23 2.4 NOISE REDUCTION TECHNIQUES IN EEG.......................................................... 24 2.4.1 Classical Techniques ............................................................................ 24 2.4.2 Machine Learning Techniques ............................................................. 29 2.5 CONVOLUTIONAL AUTOENCODERS .................................................................. 32 2.5.1 Examples of CAE noise reduction ........................................................ 33 2.6 SUMMARY ........................................................................................................ 34 2.6.1 Overview ............................................................................................... 34 iv 2.6.2 Gaps in Literature ................................................................................ 35 2.6.3 Research Question ................................................................................ 35 3 EXPERIMENT, DESIGN & METHODOLOGY ............................................ 36 3.1 HYPOTHESIS ..................................................................................................... 36 3.2 DATA ................................................................................................................ 37 3.2.1 Data Collection ..................................................................................... 37 3.2.2 Data Preparation .................................................................................. 38 3.3 CONVOLUTIONAL AUTOENCODER DESIGN ....................................................... 40 3.3.1 Hyperparameter Tuning ....................................................................... 40 3.3.2 Final Model Selection ........................................................................... 42 3.4 EVALUATION OF RESULTS ................................................................................ 44 3.4.1 Signal-to-noise ratio ............................................................................. 44 3.4.2 Hypothesis Testing ................................................................................ 45 3.5 SUMMARY ........................................................................................................ 48 3.5.1 Strengths ............................................................................................... 49 3.5.2 Limitations ............................................................................................ 51 4 RESULTS, EVALUATION & DISCUSSION ................................................. 54 4.1 DATA EVALUATION .......................................................................................... 54 4.2 CAE MODEL SELECTION .................................................................................. 55 4.2.1 Hyperparameter Selection .................................................................... 55 4.2.2 Final Model Accuracy, Reliability, Efficiency ...................................... 57 4.3 EXPERIMENT RESULTS ..................................................................................... 60 4.3.1 Single Stacked Convolutional Autoencoder.......................................... 61 4.3.2 Multiple Stacked Convolutional Autoencoder ...................................... 68 4.4 EVALUATION OF RESULTS ................................................................................ 72 4.4.1 Single Stacked Convolutional