Castañeda Tabares, Nichool
Loading...
1 results
Publication Search Results
Now showing 1 - 1 of 1
Publication Time frequency analysis and deep neural networks for visual and auditory working memory load classification from electroencephalogram(2024-05-08) Castañeda Tabares, Nichool; Manian, Vidya; College of Arts and Sciences - Art; Bustillo Zarate, Alcibiades; Portnoy, Arturo; Department of Mathematics; Lugo Beauchamp, Wilfredo E.Electroencephalogram (EEG) is a high-precision instrument to analyze cognitive processes. In this work, sixteen EEG channels are used to study auditory and visual working memory load in healthy people. With this objective, we recruited fourteen volunteers; in four of them, we evaluated the auditory working memory load and the visual working memory load in the remaining ten. Then, each of the attempts made by the participants was assigned a class: class 0 or class 1, depending on the precision they obtained with the stimuli. With the data collected, feature extraction tools such as the correlation matrix with Pearson coefficients and the Empirical Decomposition in Ensemble Modes algorithm (EEMD) were applied. Then, the Independent Component Analysis (ICA) process was carried out to select the five most relevant characteristics. Subsequently, intending to determine a person’s accuracy when faced with a working memory load stimulus, we implemented a deep neural network. In this process, results were obtained that imply that the deep neural network correctly classifies more than 90% of the attempts. We observed that the classification accuracy for the auditory working memory load was 99.91%, implying that the model almost perfectly classifies the test data. For the visual working memory load, the accuracy was 93.82%, with very good classification performance. Therefore, we can conclude that the methodology used to evaluate auditory and visual working memory load is ideal for future research and applications due to its high precision.