Klasifikace signálu EEG - KNN klasifikátorem

Abstract

The aim of the thesis is EEG classification. Evaluation of the EEG recordings is very demanding process without any software. For this reason, several classifiers able to extract the significant features have been developed. After the theoretical analysis, a k-Nearest Neighbors (k-NN) algorithm was chosen as the method to be used in this thesis since it is suitable for the cases where no information about the data set is known. In the practical part, the k-NN classifier is implemented using Matlab. The EEG recording is divided into quasi-stationary segments which ensures that the signal can be described by the features suitable for automatic classification. The prototype utilized for the classification is determined by using the c-means clustering. The results were verified by comparing with the Brain Quick program.