Brain computer interface using detection of movement intention

For the patient with extensive paralysis, developments in the emerging area of Brain Computer Interfaces (BCI) offer the prospect that some level of independent function can be regained by using signals recorded from the brain to control mobility aids or other forms of assistive technology. However, current BCI's are often slow and require extended user training. There is also a lack of multi-dimensional control and this places a limitation on the type of assistive technology that can be used with a BCI.The current study aims to investigate whether multidimensional control can be achieved from classification of brain activity recorded by surface electro-encephalogram (EEG) that precedes movement during motor tasks associated with rapid point-to-point movements of the wrist in different directions (or of the imagination of this task). The hypothesis is that because of the known properties of cortical neurones from the different areas of the cortex the electrical fields associated with this type of movement will be classifiable in relation to direction of movement of the wrist. Experiments were conducted with local ethical approval on normal subjects and EEG data from high density electrode montages were recorded.The study successfully identified the existence of statistical differences in the relative power of the EEG in the alpha, beta and gamma bands during the time preceding movement initiation related to movement in different directions. Classification of single pre-movement EEG epochs based on Euclidean Distance, K-nearest neighbour and binary decision tree techniques resulted in high success rates of upto 95%. These classification results support the hypothesis that the production or imagination of rapid wrist movements to different directions can be used for robust BCI systems that based on these results could achieved 4 separable command states. The construction of topographic maps of the success rates achieved in these classification results also reveal considerable variability in the electrode sites that produce the highest classification rates and this highlights the need for careful consideration of the number and location of EEG recording sites needed for multi-dimensional BCI systems. The work completed in this thesis has demonstrated that multi-dimensional control can be achieved by EEG based BCI devices that do not require computationally expensive algorithms for intention detection and classification.

Name dc:type.qualificationname

phd

Level dc:type.qualificationlevel

doctoral-pg

Grantor dc:publisher.institution

University of Strathclyde

Year dc:date.issued

2007

Author dc:creator

• Valsan, Gopal

Advisor dc:contributor.advisor

• Conway, B. A.

DOI

10.48730/5414-zx85

Identifier

uk.bl.ethos.442007
T11847

OAI identifier oai:identifier

oai:strathclyde:m900nt420