Özet:

Electroencephalogram (EEG) signals are mostly used to measure mental workload for reasons such as portability, practicality of use and ease of signal reception. Extraction and classification of features from EEG signals is an important issue. In this study, the features created by the power spectral density values of the EEG signals are classified by deep learning and machine learning methods. It has been shown that the results obtained with the deep learning algorithm are superior to the classical machine learning algorithms. Firstly, the participants were given arithmetic tasks such as subtracting two different numbers from each other mathematically. The EEG signals of the participants at rest and during the arithmetic task were recorded. Subbands were obtained by applying discrete wavelet transform (DWT) to EEG signals. The power values of the bands were calculated with the Welch method. Feature vectors were created by combining the power values from each band. These feature vectors were classified using different machine learning algorithms such as long short-term memory (LSTM) deep learning algorithm and support vector machine (SVM), linear discriminant analysis (LDA), k-Nearest Neighborhood (k-NN). As a result of the classification, the performance metrics such as accuracy, sensitivity, specificity, F1 score and precision calculated for each channel were averaged and the superioty of the classification algorithms were compared in table.

Anahtar Kelimeler: