Brain computer interface signals increased accuracy with classification of different time segments

Recent studies in biomedical engineering and neurology showed that a healthy human brain can control a variety of electronic devices only by thinking without making any muscle movements, in this way it can be created brain computer interface (BCI) systems that facilitate the lives of paralyzed patients. In this thesis, in order to facilitate the lives of patients who can not use muscle systems such as amyotrophic lateral sclerosis, electroencephalography (EEG) and electrocorticography (ECoG) signals were used in BCI systems. Then, It aimed to increase the classification accuracy with features to be extracted from different time periods (epochs) of these signals. In this thesis study, the following three data sets are used for the creation of faster and more accurate BCI systems: BCI Competition 2003 Data Set Ia (Data Set 1), BCI Competition 2005 Data Set I (Data Set 2) and EEG data set recorded from 3 healthy individuals taking necessary ethical permissions in the EEG Research Laboratory of the Electrical and Electronics Engineering Department of the Karadeniz Technical University (Data Set 3). The results obtained are compared with the literature studies, then performance of this method used in the thesis was evaluated. According to these results, 99.31% and 99.00% classification accuracy (SD) were calculated with Data Set 1 and Data Set 2 which the best results in the literature. For Data Set 3, 82.24%, 62.50%, 57.23% SD were calculated for A, B and C respectively. According to the studies in the literature, SD was increased by 4.6% and 1.31% for B and C subjects, whereas SD did not change for A subject.