This work consists in developing software in a microcontroller which could be applied in the control of a brainmachine interface. Parameters from the pre-processing and processing were replicated from a control developed by Alexandre Ormiga Galvão Barbosa in his Master s Thesis, which was developed in an off-board capacity, using a computer. The software developed is intended to identify thought patterns of a human mind using electroencephalograpy (EEG) readings. A data base of EEG signals with known outputs was used to develop an artificial neural network capable of identifying these signals. A block of samples refering to a mental activity is submitted through a series of preprocessings and adapted into one input for the neural network, which outputs an attempt to indetify the pattern as one of three possibilities. The data base was separated in a training set for the neural network and another base for test. Software was developed in Python capable of applying the desired pre-processings to the training set and then apply them in training the neural network. Finally, a circuit was implemented to test the developed neural network using a second microcontroller to supply the trained neural network loaded in a microcontroller with train set data, simulating reading an EEG.