| Resumen |
In this work, a new experimental model of feature extraction for EEG signals is proposed, having as main characteristic the generation of patterns that can be classified with less computational cost and better performance. The proposed algorithm is based on segmentation stripes of the signal applied to a test consisting on recalling the name of one color out of 4, that are presented to a subject of study in screen. During this period the EEG activity of the subject is measured with an EMOTIV device. The proposed segmentation method was applied to 100 samples to classify them according to the corresponding color, and then 9 different machine learning techniques were used. In 8 of them, the proposed model performed better; the simplicity of our model helps to achieve better performance in space and time at the time of classification. |
