EXPLORING CONVOLUTION FOR THE DETECTION OF MENTAL ATTENTION STATES

Recent technological advancements have resulted in innovative operational environments in which human actors are increasingly relegated to passive observers. While such settings open up new horizons, as well as lifestyle advances, they pose risks associated with humans’ incapacity to retain focus and concentration during passive control tasks. As machine learning models become more sophisticated and biometric data becomes more readily available through new non-invasive technologies, it becomes increasingly possible to gain access to data that could revolutionize Brain Computer Interface (BCI) systems. In this study, the feasibility of a passive BCI system relying on electroencephalography (EEG) to distinguish mental attention states is investigated. The data used is acquired from five participants engaged in a computer-simulated train control task. The aim is to build an infrastructure that would rely on a consumer-grade device to distinguish between mental states, namely focus, detachment (or disengagement) and drowsiness. Classifiers such as Linear Discriminant Analysis (LDA), Support Vector Machines (SVM) and K-Nearest Neighbors (KNN) are used in this work and serve as a beneficial baseline. Moreover, because of the promising preliminary results, artificial neural networks (ANN) are explored and their full potential is reached when convolution is used. Convolutional neural networks (CNN) are used to develop both a subject-independent and a subject-specific classifier, as well as to apply transfer learning. The findings provide a ground basis for further research and may help guide the design of future systems for monitoring the state of humans by means of EEG brain activity data.