Automatic detection of mental workload from eeg signals using signal discrimination techniques and deep learning model

Mental workload is the amount of mental work required to complete a task over a given period of time. EEG signals are mostly used to measure mental workload for reasons such as portability, practicality of use and ease of signal reception. In the literature, many studies have been carried out using different feature extraction and classification algorithms related to the classification of mental workload. However, limited success has been achieved in these studies. In this thesis, the participants were given the arithmetic task of serially subtracting two different numbers. The EEG signals of the participants in their resting state and during the arithmetic task were recorded using an EEG device with 23 channels and a sampling frequency of 500 Hz. By applying DWT and EMD to raw EEG signals, delta (0-4 hz), theta (4-8 hz), alpha (8-16 hz) and beta (16-32 hz) bands were obtained. PSD values of the bands were calculated using the Welch method. The feature vectors created with DWT+Welch, EMD+Welch and Welch were classified using LSTM deep learning algorithm and SVM and k-NN machine learning algorithms. As a result of the classification, the performance metrics of each classifier are shown in charts on the basis of channels.