Automated EEG signal processing: A comprehensive investigation into preprocessing techniques and sub-band extraction for enhanced brain-computer interface applications

Abstract

The Electroencephalogram (EEG) is a vital physiological signal for monitoring brain activity and understanding neurological capacities, disabilities, and cognitive processes. Analyzing and classifying EEG signals are key to assessing an individual’s reactions to various stimuli. Manual EEG analysis is time-consuming and labor-intensive, necessitating automated tools for efficiency. Machine learning techniques often rely on preprocessing and segmentation methods to integrate automated classification into EEG signal processing, with EEG sub-band components (δ,θ,α,β and γ) playing a crucial role. This paper presents a comprehensive exploration of EEG preprocessing methods, with a specific focus on sub-band extraction techniques used in Brain-Computer Interface (BCI) applications. Various methods—including Fast Fourier Transform (FFT), Short-Time Fourier Transform (STFT), Finite Impulse Response (FIR) and Infinite Impulse Response (IIR) filters, and wavelet transforms (DWT, WPT)—are evaluated through qualitative and quantitative parametric analysis, along with a review of their practical applicability. The study also includes an application-based evaluation using an open-access EEG dataset for drowsiness detection.