EEG Signal Processing and Feature Extraction: A Review of Traditional and Deep Learning Frameworks

Electroencephalogram (EEG) represents a fundamental non-invasive neuroimaging modality for recording dynamic cortical activity. Despite its extensive clinical and technological applications, raw EEG signals are highly susceptible to physiological artifacts and environmental noise, necessitating robust signal conditioning pipelines. This comprehensive review systematically analyzes the complete EEG signal processing continuum, encompassing data acquisition, preprocessing, feature extraction, and post-processing refinement. We initially evaluate traditional mathematical methodologies, including Independent Component Analysis (ICA), discrete wavelet transforms (DWT), and Principal Component Analysis (PCA), which provide foundational mechanisms for noise suppression, spectral feature representation, and dimensionality reduction. However, their reliance on manual feature engineering limits adaptability across complex, non-stationary time series. To overcome these limitations, this paper explores the transformative integration of Machine Learning (ML) and Deep Learning (DL) paradigms within the processing workflow. We highlight how intelligent architectures, such as Support Vector Machines (SVM) and Random Forests (RF) for automated artifact classification, Genetic Algorithms (GA) for feature selection, and Convolutional Neural Networks (CNN) combined with Long Short-Term Memory (LSTM) networks for end-to-end spatiotemporal feature learning, significantly enhance signal quality and decoding accuracy. Furthermore, we explore emerging trends, including Graph Signal Processing (GSP), Attention Mechanisms, and Federated Learning (FL), to model complex functional connectivity and ensure data privacy. Ultimately, this review presents a unified roadmap illustrating the evolution from classical signal conditioning to advanced artificial intelligence frameworks.

کلیدواژه ها:

Electroencephalogram ، Signal Processing ، Machine Learning ، Deep Learning ، Feature Extraction ، Artifact Removal ، Brain Computer Interface