Classifying Known/Unknown Information in the Brain using Electroencephalography (EEG) Signal Analysis

Electroencephalography (EEG) was studied as another method for detecting in addition to existing polygraph tools, on the premise that EEG signals are more difficult to fool than physiological responses to polygraphs. In the interrogation process carried out by the authorities with the aim of extracting memories and information whether the suspect or witness knows or does not know the object being clarified. But the suspect will also try to cover up the truth by lying and deception. This research is focused on finding spatial patterns of statistical parameters by analyzing EEG signals when the defendant or witness is given a stimulus object in the form of an image. The EEG channels used were T3, T4, T5, T6, 01, and O2. The six channels are analyzed for the alpha, beta, and gamma EEG-subband to find the average of mean, MAV and STD values which will be used as parameters to classify spatial known/unknown objects in the image as a stimulus. The results of this study indicate that there is a pattern where the EEG features in unknown conditions tend to be higher compared to the known condition. Furthermore, the EEG signal data is classified using 4 machine learning algorithms namely Naïve Bayes, SVM, KNN, and Neural Network. Thus, the optimum result was obtained by the KNN algorithm with 87% of accuracy.