| Summary: | CognitiveWorkload (CW) is a key factor in the human learning context. Knowing the optimal amount of CW is essential to maximise cognitive performance, emerging as an important variable in e-learning systems and Brain-Computer Interfaces (BCI) applications. Functional Near-Infrared Spectroscopy (fNIRS) has emerged as a promising avenue of brain discovery because of its easy setup and robust results. It is, in fact, along with Electroencephalography (EEG), an encouraging technique in the context of BCI. Brain- Computer Interfaces, by tracking the user’s cognitive state, are suitable for educational systems. Thus, this work sought to validate the fNIRS technique for monitoring different CW stages. For this purpose, we acquired the fNIRS and EEG signals when performing cognitive tasks, which included a progressive increase of difficulty and simulation of the learning process. We also used the breathing sensor and the participants’ facial expressions to assess their cognitive status. We found that both visual inspections of fNIRS signals and power spectral analysis of EEG bands are not sufficient for discriminating cognitive states, nor quantify CW. However, by applying machine learning (ML) algorithms, we were able to distinguish these states with mean accuracies of 79.8%, reaching a value of 100% in one specific case. Our findings provide evidence that fNIRS technique has the potential to monitor different levels of CW. Furthermore, our results suggest that this technique allied with the EEG and combined via ML algorithms is a promising tool to be used in the e-learning and BCI fields for its skill to discriminate and characterize cognitive states.
|
|---|