Classification of sleep disturbance and deep sleep using FFT, PCA, and neural network

Abstract

This paper proposes a method to classify sleep disturbance and deep sleep using electroencephalogram (EEG) signals at sleep stage 2, fast Fourier transforms (FFT), and principal component analysis (PCA). In order to extract the initial features, the FFT was carried out to remove noise from EEG signals at sleep stage 2 in the first step. In the second step, the noise-free EEG signal extracted in the first step was reduced to five dimensions using the PCA. In the final step, the classification performance was measured using the five dimensions as input to a neural network with weighted fuzzy membership functions (NEWFM). In classification performance, accuracy, specificity, and sensitivity were all 100%.