Publicação
A comparison between geometric properties and central moments to detect P300 waves
| datacite.subject.fos | Departamento de Informática | pt_PT |
| dc.contributor.advisor | Fonseca, Manuel João Caneira Monteiro da, 1968- | |
| dc.contributor.author | Cardoso, João Ricardo Dias | |
| dc.date.accessioned | 2018-11-06T16:41:22Z | |
| dc.date.available | 2018-11-06T16:41:22Z | |
| dc.date.issued | 2018 | |
| dc.date.submitted | 2018 | |
| dc.description | Tese de mestrado, Informática, Universidade de Lisboa, Faculdade de Ciências, 2018 | pt_PT |
| dc.description.abstract | Brain-Computer Interfaces (BCI) are a way to communicate without using any mus- cle movement, using only signals generated by the brain and collected using Electroen- cephalogram (EEG). This kind of applications are appropriate for people with physical disabilities since they cannot use devices like the mouse or the keyboard. One of the paradigms of the BCI applications is the P300. This is a signal that happens when we identify or recognize something that we are waiting for. A practical application of these BCIs are the Spellers that contain letters and allow users to write. The Spellers light the letters randomly, leading to the generation of P300 signals when the desired letters are highlighted. There are several methods for detecting the P300, but most of them require training and calibration prior to use to achieve acceptable success rates. Some even need to be calibrated for each user before they can be used. With this work we intend to develop two new methods to detect the P300 signal and compare them to find the best. The first one uses physical features of the signal (geometric shape) and the second uses regions of the signals, described with central moments. For both methods we intend that they do not need individual training. To do this, we studied the existing approaches to detect P300, and analyzed some Spellers. For the creation of the first method, we described the signals using a set of geometric properties. We also conducted tests to find the best classifier and created an EEG signal pre-processing pipeline allowing our method to use signals from different record devices. For the creation of the second method we conducted the same steps, however we chose a set of regions of the signal to describe the signal and in each of these regions we used central moments to describe them. Finally, we conducted an exper- imental evaluation to compare our methods with others. The results showed that between our methods the best one is the central moments method, since it showed in almost all users accuracies above 90% for 2 datasets. However, the geometric models had close ac- curacies but not enough to overtake the central moments model. In the last dataset, from which we had the accuracy of the Stepwise Linear Discriminant Analysis (SWLDA) from the authors, none of our methods had an average accuracy value above 80%. However, the central moments model, presented results above 80% for two users and in the rest of the users presented accuracy values close to the results of the SWLDA. | pt_PT |
| dc.identifier.tid | 202011208 | pt_PT |
| dc.identifier.uri | http://hdl.handle.net/10451/35288 | |
| dc.language.iso | eng | pt_PT |
| dc.subject | Interface Cérebro-Computador | pt_PT |
| dc.subject | P300 | pt_PT |
| dc.subject | Deteção de P300 | pt_PT |
| dc.subject | Propriedades geométricas | pt_PT |
| dc.subject | Momentos centrais | pt_PT |
| dc.subject | Teses de mestrado - 2018 | pt_PT |
| dc.title | A comparison between geometric properties and central moments to detect P300 waves | pt_PT |
| dc.type | master thesis | |
| dspace.entity.type | Publication | |
| rcaap.rights | openAccess | pt_PT |
| rcaap.type | masterThesis | pt_PT |
| thesis.degree.name | Mestrado em Informática | pt_PT |