A Pilot Investigation on the Performance of Auditory Stimuli based on EEG Signals Classification for BCI Applications

Abstract

Brain Computer Interface (BCI) is a communication pathway between the external devices and the brain signals that doesn’t require any physical activity of the muscular system. Such systems are the only mode of communication for people affected by a number of motor disabilities. In some medical conditions, the person is conscious and awake, but all of his voluntary muscles are paralyzed. Some patients retain vertical eye movement and partially recover from the muscular paralysis. For such patients, there are numerous communication devices available in the market. But for patients who are affected completely, that means the eyes, as well as the muscular activity, is completely paralyzed, hearing is the only mode for them to communicate. In this paper, various auditory stimuli are explored that can be used in BCI applications. In order to create an auditory simulation, comforting sounds to the users such as music and natural sounds are used. This work uses 6 different sounds as auditory stimuli and the brain signals are recorded using an electroencephalogram. The auditory signals are further classified with various classification algorithms such as multi-layer perceptron, random forest, and decision trees. The performance has been analyzed in terms of accuracy, precision, and recall. The average accuracy of 91.56% has been obtained for the random forest, 86.78% for decision trees and 89.92% for multi-layer perceptron. Random forest shows the best classification accuracy when compared to other two classifiers while classifying auditory stimuli-based EEG signals.