Abstract
The purpose of this study was to apply the EEG to BMI in order to realize appropriate control of the classification of subjects' intentions. The recorded EEGs were additive averaged and then compared with the EEGs of the left and right legs. The EEG recordings were additionally averaged, followed by ANOVA and multiple comparisons based on RP peaks, and classification of movement conditions using LSTM. The additive averaging revealed typical RP waveforms. Significant differences in the peak amplitude of RP were observed between the movement conditions at the frontal lobe measurement site. The mean values were larger in the order of condition I, M, and S, and the confidence intervals differed among the stimulus conditions. Furthermore, although the dominant foot of all subjects was the right, the mean value was larger for the left foot. Since the prefrontal cortex measures judgment and thinking, the mean value was also larger in the condition with stronger prefrontal cortex activity, suggesting a significant difference. LSTM was able to classify between conditions with high accuracy, 97.0% accuracy and 0.11% loss in the final value.
