Information Technology Reference
In-Depth Information
Fig. 2.
Example of EEG map. The scale is normalized between 0 and 1 as it can be
seen on the scale bar. Each electrode is placed in its particular position and the value
generates the map.
3
Classification of Mental Tasks Using Computer Vision
After defining the protocol to obtain the EEG mapping images, a further image
processing must be developed to obtain the mental tasks.
The technique proposed is based on a general image or pattern for each mental
task and user that can be compared with each trial of data individually to
determine which mental task is performing the user. The model of each task are
obtained for each user through a training that consists of an oine session of
several minutes. This data are processed in trials of 5 seconds that are averaged
to obtain a final model for each mental task and used in the classifier.
First, the significant frequencies (the ones where the difference is substan-
tial) have been chosen using a qualitative approach. Afterwards, a classification
algorithm is proposed to differentiate between the different mental tasks.
3.1 Frequency Choice
As it was mentioned in Section 2, 12 different frequencies have been obtained
after preprocessing the raw EEG signals. To classify the different mental tasks
it is essential to have an appreciable difference between the maps obtained for
each tasks. To that end, a qualitative analysis of the data of each subject has
been made.
The most significant change between tasks appears in frequencies between
8-14 Hz. In particular, 12 Hz is the most significant frequency for Subject 1, 10
Hz for Subject 2 and, finally, 14 Hz for Subject 3. As it is shown in Figure 3,
different shapes on the image can be clearly seen for left, right and word mental
taks in the EEG mapping for these frequencies.
Search WWH ::
Custom Search