Identificação de emoções básicas de seres humanos por meio do processamento de sinais de eletroencefalograma

Abstract

The human brain is capable of emitting waves at certain frequencies, which can be detected with an electroencephalogram (EEG) device. With the recording of those waves, it is posible to identify one’s emotions. In this work it is shown the procedure involved on EEG signal aquisition, the processing of this signal and its classification, done by a machine learning algorithm which has the goal of identfying basic emotions. This medical procedure can also help in the diagnosis of neurological diseases such as epilepsy and Parkinson’s disease. In addition, the identification of emotions can also help in the diagnosis of psychiatric illnesses, such as sleep disorders, alcohol and drug abuse and others. The EEG device captures the potential difference between the electrodes coupled to the individual’s scalp, graphs are generated with these potential differences over time. The recorded signals then undergo pre-processing to attenuate noise and, subsequently, the machine learning algorithms Ensemble KNN, Quadractic SVM and Decision Tree will perform the classification. With the implementation of these processes, it was possible to obtain up to 98,57% accuracy rate using Ensemble KNN.