Detection of Auditory Selective Attention Using Artificial Neural Networks: An Intersubject Analysis

Abstract

The auditory selective attention is the ability that allows the concentration on a sound stimulus of interest while ignoring other stimuli. The classification of this attention state might be done through auditory steady-state responses, being a possible application in brain-computer interfaces. A method to perform the classification of selective attention is proposed in this article, with dimensionality reduction by principal component analysis, filtering of the signals by a digital Butterworth filter and the computation of the energies of the resultant signals. The energy values are then applied to the inputs of an artificial neural network to perform the classification, obtaining a max accuracy of 64.07% with an information transfer rate of 2.7197 bits/min. So, it is shown that the classification of the effect is possible, however it is still necessary some studies to tell how much the performance of this classification can be improved.