| dc.description.abstract | Currently, Functional Near-Infrared Spectroscopy (fNIRS), which is the non-invasive method, is one of the most effective techniques for brain activity analysis. The successive changes in real time from concentration of the two principal parameters including blood oxyhemoglobin (Oxy-Hb) and de-oxygenated hemoglobin (Deoxy-Hb) are measured using the fNIRS-FOIRE 3000 system. This study focuses on binary classification betweeen resk and work task to investigate the mental workload of subjects during neurocognitive training task Stroop Task experiment four classification algorithms.
The experiment retrieve data from 20 healthy subjects whose aged from 18-21 involved in the Stroop Task experiment. The accuracy and reaction time results are compared to show the difference in difficulty between two levels of Stroop task: Inverse and Conventional. After that, the data is the input of four classification method including k nearest neighbor (kNN), support vector machine (SVM), artificial neural network (ANN) and randomized decision forest implemented in Python language. The data is then compared in three ways: classification type, features and classification algorithms. The results show that binary classification is more effective than multiclass classification while decision tree combined with skewness resulted in the higest accuracies with 92%. However, generally statistical features and randomized decision forest claimed their effectiveness and reliability in discriminative mental workload classfication. | en_US |
