Classification of Cognitive Load Based on Neurophysiological Features From Functional Near-Infrared Spectroscopy and Electrocardiography Signals on n-Back Task

Abstract

Cognitive load can be estimated using individuals' task performance, their subjective measures, and neurophysiological measures. Neurophysiological measures, which among others include brain activation signals obtained with various brain imaging techniques, such as the functional near-infrared spectroscopy (fNIRS), and signals from the peripheral physiology, such as the electrocardiography (ECG) signal, allow an objective and continuous estimation of cognitive load. In this article, the fNIRS and ECG signals were simultaneously collected from 32 participants and used to classify three levels of cognitive load on n-back task. A set of 30 fNIRS and ECG features proposed in this article enables the classification of different levels of cognitive load on n-back task using the support vector machine (SVM), k-nearest neighbors (KNN), and linear discriminant analysis (LDA) classification models. When combining the fNIRS and ECG features, three difficulties of the n-back task were classified with the mean accuracies ranging from 61% to 67%, while two difficulties were classified with the mean accuracy ranging from 70% to 84%. The most important features in the classification are discussed. The results presented in this article extend the existing empirical evidence that combining brain imaging and peripheral physiology features increases the accuracy of multi-level cognitive load classification, thus further underscoring the importance of multimodal approach to cognitive load classification.