In this thesis two different studies are presented separately because each study has a different purpose even though the two studies are complementary. For the first study an available dataset for BCI is used. This dataset is consistent with the aim of the first study because it allows the implementation and development of the proposed technique for BCI based on imaginary limb movement with a small amount of electrodes. The technique I implemented trough this study is a reliable option to be used for a BCI-based on imaginary movements. In the second study a different dataset is used because the requirements for the technique to be implemented are different. Training and testing dataset are necessary to describe a practical computational technique that is applied for a single trial classification based on a Common Spatial Pattern and complemented with a linear discriminator. The training data is used to build a classifier and the testing data allows measuring the performance of the classifier implemented. After testing the classifier the good performance obtained makes this technique of great interest for any imaginary movement based-BCI.