A Review of Electroencephalography (EEG) Signal Processing Methods: From Preprocessing to Deep Learning

Javid Eskandarifard¹ Afsaneh Najafi²

1) Food and Drug Administration (FDA), Tehran, Iran, Email:
2) Islamic Azad University, Tehran Medical Sciences Branch, Tehran, Iran, Email:

IOI Code : ACCD-ACAABA
Publication : 10th International Conference on Applied Researches in Science & Engineering (carseconf.com)
Abstract :
    Electroencephalography (EEG) is a non-invasive technique for recording the brain’s electrical activity and is characterized by high temporal resolution. Despite this advantage, the low amplitude and temporal variability of EEG signals pose significant challenges for their analysis. Eye blinks, muscle activity, electrode movement, and power-line interference can also distort neural information. This review examines the major stages of EEG signal processing, from quality control and preprocessing to artifact removal (the elimination of unwanted noise from recorded signals), time–frequency analysis, feature extraction, and model evaluation. Classical signal processing techniques are discussed alongside machine learning and deep learning algorithms, graph neural networks, and self-supervised learning approaches. The evidence indicates that no single method is universally suitable for all applications and datasets. The selection of processing steps should be aligned with the physiological objective, electrode configuration, noise characteristics, and operational conditions of the system. Furthermore, randomly splitting overlapping windows between training and test sets may lead to data leakage and unrealistic estimates of model performance. Subject-level or session-level validation, together with an assessment of neural component preservation, is essential for reliable evaluation. Accordingly, this review concludes by presenting a practical framework for designing and reporting EEG signal processing pipelines.

Keywords : electroencephalography (eeg) signal processing artifact removal time frequency analysis deep learning brain computer interface (bci)