Motor Imagery EEG Classification with the SNN-based NeuCube Framework

Shumska, Mariya (2021) Motor Imagery EEG Classification with the SNN-based NeuCube Framework. Bachelor's Thesis, Computing Science.

Preview	Text bCS_2021_ShumskaM.pdf Download (4MB) | Preview
	Text toestemming.pdf Restricted to Registered users only Download (138kB)

Abstract

A brain-computer interface (BCI) bridges the gap in communication between humans and computers by analysing brain activity and controlling external devices via motor imagery. BCIs typically use the brain signals recorded from the scalp surface (EEG). Recently, the Deep Learning (DL) techniques proved to be very accurate classifiers of users’ intents. However, the bottleneck of DL frameworks is high power and memory consumption which might be problematic for portable devices with BCI. The list of such devices includes medical technologies for motor rehabilitation, cognitive disorders treatment, gadgets for games, or smart environment control. Inspired by the progress of Spiking Neural Networks (SNNs) in pattern recognition, where they demonstrated excellent performance in speech recognition, visual processing, and medical diagnosis, we would like to explore their potential for EEG motor imagery classification. We conduct an experiment, where we investigate the SNN-based NeuCube framework performance on raw and preprocessed data set. Preprocessing methods include filtering and Independent Component Analysis (ICA). We compare several ICA algorithms: Infomax, FastICA, and FastICA with prior dimensionality reduction with PCA. In summary, our experiment showed that preprocessing, especially FastICA-based methods, improves the results. However, the accuracy obtained with the NeuCube framework is lower in comparison to other SNN-based classifiers used on the same data set.

Item Type:	Thesis (Bachelor's Thesis)
Supervisor name:	Bunte, K. and Biehl, M.
Degree programme:	Computing Science
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	02 Aug 2021 09:42
Last Modified:	02 Aug 2021 09:42
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/25555

Actions (login required)

View Item