| Abstract [eng] |
In this project it is described brain-computer interface design and implementation. Model designed using empirical data, because human brain‘s physiological activity is yet not sufficiently investigated. The model designed for one task. Described model is designed for lamp control, turn it on or off. Brain activity related to lamp control is noted as source signal, all the other brain activity is noted as noise signal. Suggested system uses 3 electrodes, whereas usual systems use 20 or more electrodes. It is described the created methology for conducting the experiment and getting learning data for model. In experiments the lamp control is simulated by means of keyboard presses. Brain activity signal and keyboard key presses are captured and synchronized in time. Two programs Matlab and openvibe are compared by means of data processing. Data processing is carried by band filter and data division into intervals. Comparing of two classifiers is conducted: Linear discriminant analysis and support vector machine. To determine trainings error dependence, the following experiments were performed: changing training time, changing time window between lamp control on or off, changing placement of electrodes, changing number of controlled lamps. |
