EEG Signal Processing

After recording electroencephalogram (EEG) signals based on the research purpose, it is essential to examine the signal content for the information contained therein. The information provided by an EEG is not necessarily in the realm of time and seemingly random fluctuations. In fact, we should look for a feature in the studied signal that properly represents the changes made at each stage of the experiment. Such features in an EEG signal at rest usually include the domain of frequency power (Fourier transform), coherence (paired communication between channel signals) in time and frequency, time-frequency values (Violet Conversion, short-time Fourier transform), and the high variety of machine learning algorithms. The selection of the right solution to the processing of signals depends on our definition of the EEG recording experiment.


The features used to summarize ERP (event-related potential) signals also include a variety of methods in the time domain (single-trial averaging) and finding the minimum and maximum point at specific times, which are highly popular and reputable particularly among psychologists and psychiatrists. Nevertheless, other methods such as machine learning algorithms can also be employed to find features that properly represent the signal. After finding the most appropriate feature to represent the EEG signal, the next stage is usually to develop classifications by using a variety of classifiers. The most common classifiers used in this domain include SVM, KNN, and LDA, which can be also expanded to a variety of other methods.


Medical screening is one of the most important neuroscience projects defined based on electroencephalogram signals. This is usually performed by using specific tasks during ERP signaling and then selecting the appropriate features and categories for the removal of the intended biomarker. Continuous EEG signals at rest can also be used for this purpose. Another application of EEG signaling is to find the brain correlates based on features derived from psychometric and behavioral questionnaires. The main objective of such studies is to indirectly obtain a biomarker. To this end, correlates of brain signals are first found and then their relationship with various dimensions derived from the questionnaire is investigated by using the cross-correlation analysis.


MATLAB-based EEGLAB is one of the most common tools used for data analysis at VNRC, especially in the preprocessing stage and for the representation of final features obtained from brain signals. However, there are various plugins that can run concurrently with EEGLAB and allow researchers to apply many tests on the recorded data.

Signal Processing Services 

Data preprocessing (including the removal of biological artifacts and non-biological noises), extraction of some features such as power, asymmetry, and coherence in sensor space as well as resources (mainly through LORETA and other methods based on minimum-norm) from the data at rest, and ERP are performed in this center.