Functional Data Analysis for Decoding Working Memory Representations from EEG Signals

M. Castellano, N. Acar Denizli

Electroencephalography (EEG) provides high-temporal-resolution measurements of brain activity, enabling the study of cognitive processes such as working memory (WM). A central challenge is decoding WM content from noisy, high-dimensional, and temporally structured signals. We propose a functional data analysis (FDA) framework that models EEG signals as continuous functions over time to capture their temporal dynamics. The study is based on data from 20 children performing a task involving visual, spatial, and verbal information encoding. We compare the FDA-based approach with standard classification methods commonly used in EEG analysis. Results illustrate the ability of FDA to characterize time-resolved neural patterns associated with different WM contents, providing an interpretable statistical framework for complex neurophysiological data.

Keywords: FDA, EEG, decoding

Classification and Pattern Recognition
September 2, 2026  11:20 AM
Aula 22

R. Alberich Martí, N. A. Cruz Gutierrez, R. Fernández Peralta, I. García Mosquera, A. Mir Torres, F. Rosselló Llompart

E. Junquera Álvarez, I. Díaz, B. Remeseiro

L. Muñoz Garatachea, L. Muñoz Garatachea, N. Ayuso Escuer, R. Trillo Lado