Abstract:

The aim of the experiment was to determine the accuracy of the classification of the blink as artifact that can be registered in the EEG signal. On the basis of the collected raw signal with the length of 117 seconds, fragments were identified that represent the subject’s closed eyelids and the segment when they are open. Neural networks were created and then trained using various topologies and trigger functions. A single network processed input data of 27 samples from 14 sensors of the EEG acquisition device. Six separate network characteristics were tested, which solved this classification issue with varied accuracy.