Brain-Computer Interface Based on Motor Imagery and Emotion Using Convolutional Neural Networks

Brain-Computer Interface (BCI) allows a person to move external devices using the mind without involving muscles, gestures, and other motor functions. An Electroencephalogram (EEG) is often an intermediate device in BCI. Some variables of EEG in BCI are emotion, motor imagery, focus. Some BCIs usually use one variable, while other variables are fixed. However, it is also necessary to integrate the two variables. This research used emotional variables and motor imagery without a cross-correlation factor. EEG signals as intermediate are extracted into the wave that is represented motor imagery and emotion using Wavelet transformation. Wavelet provided the Alpha, Mu, Beta, and Theta waves. We added an amplitude feature beside four waves. Then, the features are classified by Convolutional Neural Networks. The experiment was carried out using three optimization models, i.e., Adam, AdaDelta, and AdaMax, which gave the best accuracy of 90% with AdaMax and VGG16. The model classify eight classes namely “happy forward”, “happy stop”, “happy right”, “happy left”, “neutral forward”, “neutral stop”, “neutral right”, or “neutral left”.