| Summary: | The scope of the present thesis was to investigate whether a set of electroencephalography data obtained from an already conducted study was suitable for Brain Computer Interface applications. Brain Computer Interfaces are a technology that allows the direct communication between the brain and an outside device which can be either a computer or any other artificial machinery. Brain Computer Interfaces achieve so by bypassing the normal physiological way in which the brain communicates the messaging and connecting it directly to the outer device. Regarding the experimental procedure, the data were acquired from ten healthy subjects in a study that was published by Argentinean researchers where the magnitudes such as Event Related Potentials, Inter Trial Coherence and Power Spectral Density were calculated for the total number of subjects. Subjects were presented with 4 visual cues which corresponded to the classes “up”, ”down”, ”right” and “left”. The experimental procedure was repeated both for imaging and for pronouncing the prompts. What was done was to compute the same magnitudes for each separate subject and for each separate class out of the four that were presented in the initial experiment. The conditions that we compared were two: inner speech and pronounced speech. After the extraction of the diagrams, the computational procedure was repeated but this time concentrating only in the channels that corresponded to Broca’s area which is the brain area that is mainly responsible for the production of speech. We used the MNE package written in python which is a computational tool for importing and processing electroencephalography data. It possesses significant advantages since it is open source and consequently can be implemented free of charge. The outcomes of our study showed significant differences in Event Related Potentials between each one of the subject. However, for the same subject and for each one of the separate classes the differences were negligible. As far as the inter trial coherence and power diagrams were concerned, again we observed significant similarities for the four classes but for the same subject and significant differences from subject to subject. It has to be mentioned that the outcomes for the selected channels being concentrated only in Broca’s area were more or less similar to the ones for the total number of channels. Additionally to the already acquired outcomes, a further statistical analysis was conducted which consisted firstly by the computation of the topographical display of the power for each one of the subjects and for the two separate conditions but simultaneously for all of the four classes. Secondly, the time frequency plots and the power spectral density plots were computed. The final part of this statistical analysis is the computation of event related synchronizations and desynchronizations and three series of diagrams were constructed in this computational process. Firstly, the event related synchronizations and desynchronizations in bands which were computed for the four classes in sum, secondly the event related synchronizations and desynchronizations analyzed through the time frequency responses and finally the violin plots of event related synchronizations and desynchronizations. After the further evaluation of the data through the statistical analysis it has to be stated the goal of drawing the conclusion that the data are suitable for a brain computer interface application in the inner speech modality has not been fully achieved.
|
|---|
