University of Twente Student Theses

Login

Detecting the no-control state in self-paced Brain-Computer Interfaces

Stoica, Mircea (2012) Detecting the no-control state in self-paced Brain-Computer Interfaces.

[img] PDF
5MB
Abstract:The field of brain-computer interfaces (BCI) has skyrocketed in recent years, and the advent of increasingly powerful technology is rapidly bringing it to commercial use. Crucial to its adoption for many real-world applications is the ability to respond only to intentional commands of the user. This is rather difficult because of the constant and seemingly random activity of the brain. Research in BCIs based on voluntary changes of brain activity has traditionally focused on distinguishing between two or more different and reproducible patterns, for the goal of communication. Additionally distinguishing them from all other possible brain activity is a considerable challenge for the signal processing and machine learning methods commonly used. This is an exploratory study into the behavior and characteristics of different approaches to self-paced motor imagery BCIs. Subject-specific band-power features are extracted and different classifiers are applied on a publicly available dataset, consisting of four subjects performing two types of motor imagery with no cues. The usefulness of dwell times and refractory periods is also studied, and we investigate the relation between different performance metrics. Our results suggest that linear discriminant classification between the three states is the most suitable approach, given the spatial filtering methods presently available. We find that refractory periods do not generalize well to new data but that the dwell time is highly recommended. The results show that the current event-based definition of true positive rate has an inverse relation to mutual information, which may be more related to how much a subject can maintain a desired state.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Interaction Technology MSc (60030)
Link to this item:http://purl.utwente.nl/essays/62829
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page