| Popis: |
There are some limitations for the motor imagery Electroencephalogram (EEG) including low-spatial resolution, non-stationary and susceptible to noise interference. This study aims to enhance the EEG features via an visual feedback training process by improving the visual feedback training paradigm design, EEG signal acquisition and EEG signal analysis, etc. We presented a novel feature extraction method in order to systematically improve the effective of feature selection, in which the coupling properties of EEG were combined with multi-level dynamic features including the time, frequency and spatial characteristics of EEG. Four subjects were guided to perform the Motor Imagery(MI) experiment with visual feedback training. Then, the integral value, Autoregressive (AR) model coefficients, avelet energy spectrum, Common Spatial Pattern (CSP) and mutual information were calculated as indices to quantify the enhancement of EEG characteristics before and after training. Statistical analysis revealed that the classification accuracy of EEG signals after training was significantly improved. |
