Information Technology Reference
In-Depth Information
must be classified to differentiate between several cognitive processes or “mental
tasks”.
In order to distinguish between the different mental tasks, the EEG signals
should be preprocessed, the most important features of the EEG signals are
extracted, and finally these features are classified into the different classes to
obtain the final mental task. The most common classifying of the processed
signals is through different mathematical algorithms such as Linear Discriminant
Analysis (LDA), Support Vector Machine (SVM) or Bayesian Classifiers, among
others [11].
This paper studies a new approach in BCI classifiers through computer vision
techniques. The main goal is to develop a classifier to differentiate between three
different mental tasks. The features of the signals are extracted obtaining the
power spectral density (PSD) and they are represented through images that
show the frequency value in a period of time for each electrode (EEG mapping).
These images are different for each mental state, as it is explained later, and it
may be possible to classify them with a suitable image processing.
EEG mapping consists of plotting the electrical activity of the brain in a ge-
ometrical matrix. This approach gives a much more accurate and representative
view of the mental activity obtained from the electrodes placed on the scalp.
This can be done with a voltage/time representation or a frequency based rep-
resentation. Nowadays, there are several works related to brain topography to
differentiate several kinds of diagnoses, including some mental diseases whose
origin is located in EEG alterations such as epilepsy [12,13] or schizophrenia
[14]. EEG mapping has been also used in electrotherapy [15]. This kind of stud-
ies involve processing sessions of several minutes while on BCI interfaces the
frequency of each decision is critic. This work shows the results obtained for a
shorter period of processing time. To this end, the data are processed in windows
of a few seconds to obtain the EEG maps. These maps are used by the classifier
to distinguish between the three different mental tasks.
The remainder of this paper is organized as follows. In section 2, the steps
made to create the different images used in this preliminary study are shown.
Section 3 presents the classifier designed to differentiatethethreementaltasks.
In Section 4, the results obtained are shown. Finally, section 5 summarizes the
main conclusions.
2
Image Obtention Protocol
The data set V of “mental imagery, multi-class” provided by IDIAP Research
Institute for BCI Competition 2003 has been used to do the EEG mapping [16].
This data set contains data from 3 normal subjects during 4 non-feedback ses-
sions (3 for training and 1 for test). The subjects made these experiments in
4 sessions on the same day, each one lasting 4 minutes and with 5-10 minutes
breaks between them. For each session, the subjects performed three different
tasks:
 
Search WWH ::




Custom Search