Artificial Neural Networks and Their Application in EEG Signal Classification

This research shows the performance of a multilayer perceptron (MLP) neural network in the classification of electroencephalographic (EEG) signals, for which the Emotiv Insight headset is used for the signal acquisition stage in order to generate data from EEG recordings, where two types of brain patterns were previously selected for this stage: left winks and right winks; so that through preprocessing stages, feature extraction, the feature matrix of the collected data is generated and thus form the input patterns to be used by the neural network in its training process. The main objective of the research is to make use of the classification model generated as a result of the neural network training in OFF LINE mode, thus an application has been designed in which this model is tested to evaluate if the outputs it emits correspond to the type of recording that is delivered as input. Therefore, the precision performance of the model during the testing phase was 90%, thus, the model can be considered to work in an application that requires to be covered in ON LINE mode, for example, in the so-called BCI systems (Brain Computer Interface), that translate the user's brain activity into control commands.