In this project, we attempted to improve EEG signal classification results by utilizing the temporal dependency. Initially we worked with a dataset from a competition (BCI Competition IV) and we tried to find temporal dependence on this data by several methods, but all attempts have failed. Therefore, we searched for a new dataset (Brain invaders 2013a) with clear temporal dependence. First, we tried the previous method which is not based on the temporal dependency, but it did not work well (on the new dataset). We have tried to implement the solution proposed in article  and obtained a significant improvement. This solution is based on the use of the labels and therefore does not allow learning between different experiments and different subjects. At this point, we decided to set the goal of the project to be “Improving EEG Signal Classification in an unsupervised way“. We found a way to do so by using the PCA algorithm. We found a problem in using that algorithm, which we will refer to in this book, and after the project has finished, we will try to solve this problem.