Decades before it was thought, that creating mind-controlled machines and programs will be possible only in the far and „mysterious” future. Through Brain-computer interfaces (BCI) such applications are available today for everyone. The aim of BCI systems is to provide a new, non-muscular communication channel between human and machine. One of the main application-areas is to improve the quality of motor-impaired patients’ life.
In this thesis I want to introduce the most popular form of BCI systems, which is based on the technique electroencephalography (EEG). My task was to implement an offline BCI, built on the P300 event-related potential component. P300 is the manifestation of neural processes - during visually evoked meaning-association -, and is widely used by BCIs to determine subjects’ intent.
I captured signals with the so-called P300 Speller experiment. During the experiment a table with characters is generating the stimuli by flashing rows and columns randomly. If the flashed row or column contains the character, that the subject is focusing on, the P300 component can be found in his/her neural response. With the detection of P300 we can guess the assigned character, so the subject will be able to communicate only using his/her thoughts and attention.
First I set up the elements of the system used for capturing EEG, then measured and stored the brain activity of 4 subjects via the experiment. After getting acquainted with the most popular methods of the area, I implemented the signalprocessing part of the system, based on proved techniques and on own ideas.
Finally, I qualified the efficiency of the system with the measured signals, artificially-made signals and with datasets from an international competition, seen via sensitivity, specificity and accuracy. Regarding the results, the system was able to detect at least 90% of the given characters at every tested method.