Classifying Sleep Disturbance Using Sleep Stage 2 and Wavelet-Based Features

This paper classified sleep disturbance using non rapid eye movement-sleep (REM) stage 2 and a neural network with weighted fuzzy membership functions (NEWFM). In this paper, wavelet-based features using EEG signals in non-REM stage 2 were used to classify subjects who have mild difficulty falling asleep and healthy subjects. At the first phase, detail coefficients and approximation coefficients were extracted using the wavelet transform (WT) with Fpz-Cz/Pz-Oz EEG at non-REM stage 2. At the second phase, using statistical methods, including frequency distributions and the amounts of variability in frequency distributions extracted in the first stage, 40 features were extracted each from Fpz-Cz/Pz-Oz EEG. In the final phase, 80 features extracted at the second phase were used as inputs of NEWFM. In performance results, the accuracy, specificity, and sensitivity were 91.70%, 91.73%, and 91.67%, respectively.