Currently there are different devices for the acquisition of bioelectric signals, but these are limited to acquiring and depend on computers to interpret these signals, based on this problem the present research work is presented, in which the classification of two mental tasks of motor imagination in an ARM® Cortex®-based microcontroller TM4C123g, with data obtained through a brain-computer interface. The self-organizing maps of Kohonen are used for the classification of mental tasks. The data used for the training of the classifier are obtained by applying The Fast Fourier Transform to the previously obtained electroencephalographic signals. The project is divided into two stages. First, the network is trained at the 8 seconds that the training patterns last, to obtain a better-trained network. In the next stage, the network with the best performance in the microcontroller is implemented for the classification of data not used in training. The results obtained in the validation tests used by the microcontroller provides a percentage of error between 2.25%. These errors are minor or similar to reference works.