Brain Thought Recognition

In order to implement a non-invasive BCI, this project developed two deep neural networks consisting of two parts. The first one contains 10 chars while the second has 10 digits. The brain signals are represented with the help of EEG. EEG is defined as a 2D tensor with the number of time steps as the width and the number of electrodes as the height. This input is fed into a deep neural network consisting of a convolutional neural network to capture spatial dependencies and a recurrent neural network (LSTM) to find the temporal correlations of raw EEG signals. This leads to finding features that are less sensitive to variations and distortions in EEG signals to enhance the robustness of EEG classification. The classification accuracy over the two neural networks is 92%.