Background: Considering the importance of sleep for physical and mental well-being, studying brain function during sleep is crucial. One method of evaluating brain function during sleep is labeling different sleep stages and using machine learning algorithms for automatic labeling. Training classifiers requires a large amount of training data. Due to the cost of data labeling by experts, there is a need to generate signals using a simulator. The focus of this thesis is to provide a simulator for generating adult EEG signals during sleep and wakefulness. This simulator can be used for generating training data for algorithm training, testing EEG devices, educational purposes in an academic environment, and more.
Objective: The objective of this thesis is to provide a simulator for EEG signals during sleep and wakefulness to achieve goals such as reducing processing data volume, improving sleep pattern recognition in treatment centers, validating the performance of EEG recording devices, educational applications, and more. The aim of the research is to simulate EEG signals in different sleep stages, as no similar simulator has been presented in this field.
Methodology: Investigations were conducted using recorded EEG signals during sleep and wakefulness for 20 healthy males and females. Due to the presence of different frequency ranges in each sleep stage, modeling power spectral density was employed. The results of power spectral density for each sleep stage had significant differences. To address this issue, clustering algorithms were utilized. Then, using a composite Gaussian model, each power spectral density function was modeled.
Results: The selection of the best power spectral density algorithm for EEG signals, determining the number of clusters for power spectral density functions in each sleep stage using three evaluation methods, modeling power spectral density functions using the optimal number of composite Gaussians, and finally, simulating EEG signals duri