Applying machine learning and signal analysis using EPOC NeuroHeadset

OData support
Supervisor:
Gáti Kristóf
Department of Measurement and Information Systems

With the rapid evolution of technology, many devices –like BCI-s- come to trade, that only existed in the movies earlier. The EPOC NeuroHeadset belongs to the family of these appliances.

First of all, in my thesis I show the basics of the brain-computer interfaces, especially the EPOC NeuroHeadset and the necessary biological and physical background for EEG signal formation. The purposes of the paper are reviewing the steps of signal processing (e.g. ICA, DWT) with its possibilities and examining the application of machine learning in this field. Besides, the thesis demonstrates my own practical examples, it describes the complete processes and their realization from putting the device onto scalp to making the final decision.

One of the own systems is a signal recording, processing, feature vector forming and neural network based decision making process, that builds upon a paradigm between the brain and the execution of fine, non-reflexive movements. The output is a single decision about the examined test subject.

On the other hand, the other process meant to detect P300 component and analyse the capability of the device in this lately very popular field, while utilize the oddball paradigm. To do so, this system uses the algorithms, classificators (e.g. MLP, SVM), which have been suggested in the literature. The output is also a decision about the containment of target-stimuli in the given sample.

In my thesis I present the outline of the systems, the design decisions, the created supplementary modules and the details of the realization with the evaluation of the results. Finally, I share my experiences and upgrade suggestions.

Downloads

Please sign in to download the files of this thesis.