Information Technology Reference
In-Depth Information
Fig. 4.
Models Obtention Protocol
Fig. 5.
Classification Algorithm
selected as well as the significant frequency previously studied. The data of
each mental task are processed separately obtaining a collection of 5 second
images that are averaged to get a unique EEG map for each task that will
be used as the model to compare in the classifier.
- Classification (Figure 5):
after obtaining the three models (left, right
and word), the remaining set of data is used to test the classification. To
that end, the data are processed in trials of 5 seconds obtaining the EEG
map. This image is compared with the three models obtained before using
the numerical index explained above. Afterwards, the success percentage is
obtained using two modes:
1. The one with greater similarity (left, right or word) is selected one obtain-
ing directly the success percentage between the different tasks classified.
2. A threshold is used to reject the trials that can not be clearly classified
and introduces the concept of uncertainty in order to reduce the total
error.
Search WWH ::
Custom Search