Alternative methods for human-computer interaction are increasingly need\-ed, as some people are physically unable to use a computer and require an alternative type of communication. In the scope of diploma thesis, we compared two methods for classifying electroencephalogram (EEG) intervals during imagining motor activi\-ties (left and right hand grip). We used records of the EEGMMI DS database (EEG Motor Movement Imagery DataSet), which is public and freely available on the pages of the Physionet website. The methods we implemented include procedures of digital signal processing, feature extraction, machine learning, and classification. The two implemented methods are called the method using large Laplacian mask, and the Common Spatial Patterns (CSP) computational method. The obtained results showed that the CSP method is more powerful than the Laplacian mask for the selected set of EEGMMI DS database records. The average classification accuracy for the CSP method was about 59 \% while for the Laplacian mask about 57 \%. The results of the diploma thesis will contribute to better understanding the problem of extracting intervals of motor activities and finding the optimal method for translating original signals into signals in the component space, and effective classification of the intervals of imagined motor activities.
|
