Methods of Power-Band Extraction Techniques for BCI Classification

The purpose of the article is to check whether the method of estimating EEG signal energy, treated as a feature, has an impact on the classification accuracy in BCI systems. The experiments were carried out on EEG signals recorded during the imagination of left or right hand movement. For energy estimation in 8-12Hz band, the following methods were used: variance, autoregressive model, discrete Fourier transform, and mutual correlation. The usefulness of the features was checked using the 10-CV test and the LDA and QDA classifiers. The results indicate that the best average classification accuracy was obtained for bandpass filtering and the energy estimation method with the use of variance.