Implementación del algoritmo K-means para clusterización de señales EEG durante la aplicación de una prueba Stroop

Abstract

Data analysis and model generation through machine learning (Machine Learning - ML) is one of the techniques most used by the scientific community to obtain knowledge that cannot be interpreted or analyzed with the naked eye. In this document, the specific issue of data collection, processing, and modeling is addressed using clustering techniques, specifically the K-Means algorithm. The data were obtained through a brain-computer interface (ICC - BCI) tool based on electroencephalogram (EEG), while a test subject performed a Stroop-type task, which allows psychologists to evaluate people's inhibitory control. The application of this type of algorithm in this specific area seeks to identify patterns within the EEG signals related to the subject's state. Data capture was carried out at three different hours of the day, morning, evening, and night, for two consecutive days, to obtain variability in the data. Although the data sample is small, it will serve as a starting point for analyzing the K-means algorithm in EEG signals during the execution of the Stroop test. The results obtained are shown both from a technical and psychological point of view. It can be seen in the clustering carried out with the signals in the time domain that there is a clustering pattern according to the time of day in which the test was performed. On the other hand, this pattern is not so evident for the clustering with the signals in the frequency domain. From the psychological perspective, it is found that the learning and accommodation process at the time of performing a psychological test decreases its potential.