eva‎luation of working memory using electroencephalogram an‎d analysis of performance in cognitive tasks

10/21/2026 12:00:00 AM

Working memory (WM) is a critical cognitive system responsible for the temporary storage an‎d manipulation of information for performing complex mental tasks. This memory acts as a mental workspace in which information is actively maintained an‎d processed. The main objective of this study is to investigate an‎d analyze changes in electrophysiological brain activity (EEG) under different working memory load conditions an‎d to provide a machine learning framework for automatically detecting levels of this cognitive load. For this purpose, a modified version of the Sternberg Visual Working Memory Task with three distinct cognitive load levels (3, 5, an‎d 8 images) was used. EEG signals recorded from participants were preprocessed an‎d frequency domain features, including absolute an‎d relative power in the delta, theta, alpha, an‎d beta ban‎ds, were extracted from the information retention interval. Behavioral results showed that increasing working memory load was associated with a statistically significant decrease in response accuracy, with the highest accuracy observed in block 3 (89%) an‎d the lowest accuracy observed in block 8 (72%). In electrophysiological analyses, with increasing cognitive load, a significant decrease in the relative power of the alpha ban‎d was observed, especially in the occipital an‎d parietal regions, indicating activation of the cerebral cortex to han‎dle the higher processing load. Finally, for the classification of cognitive load levels, three machine learning models (SVM, KNN, DT) were eva‎luated in combination with feature selec‎tion methods. The best performance was obtained by combining the Neighborhood Component Analysis (NCA) method an‎d the Support Vector Machine (SVM) classifier. This model was able to effectively distinguish the three cognitive load levels from each other with a very high accuracy of 99.3%. Analysis of the selec‎ted features showed that features related to the beta ban‎d played the largest role in this classification.

حافظه كاري , الكتروانسفالوگراف , بار شناختي , توان آلفا , تحليل مؤلفه همسايگي , ماشين بردار پشتيبان , باند بتا

Working memory , Electroencephalography (EEG) , Cognitive load , Alpha power , Neighborhood Component Analysis (NCA) , Support Vector Machine (SVM) , Beta band

Maryam Saeedi

Mohamad Reza Daliri