Nowadays, there are millions of people around the
world suffer from the disability caused by big stroke. In recent years
we have seen a rising interest in brain computer interface (BCI)
systems that help those patients to practice their normal lives.
Therefore, this work presents a GUI application based on an offline
BCI system to test their mental capacities. This application was
designed based on three tests are alphabet, arithmetic operations and
Raven’s progressive matrices. The success of this system depends on
the choice of the processing techniques. Therefore, Discrete Wavelet
Transform (DWT) and Principal Components Analysis (PCA) were
used to extract a set of statistical features from the recorded brain
signals. These features were classified into four classes are head
movement to up, down, right or left using three classifiers are
Artificial Neural Network (ANN), Support Vector Machine (SVM)
and Linear Discriminant Analysis (LDA). The performance of
classifiers was measured using the most frequently statistical
parameters: the sensitivity, specificity, precision, classification
accuracy, and area under receiver operating characteristics (ROC)
curve (AUC). It was concluded that when DWT was used as a feature
extraction, ANN and SVM achieved the highest classification
accuracy with a value of 95.24% but when using PCA, ANN
achieved the highest classification accuracy with a value of 92.86%.
On the other hand, LDA classifier was the worst among the three
classifiers |
