Abstract—This work develops a number of automatic slowwave sleep (SWS) detection methods that employ only oneor two channels of EOG/EEG signals. In addition to thereduction of signal channels, a distinct feature of theproposed approach is the introduction of a new feature setthat can make the proposed approach insensitive tointerpersonal differences of the physiological signals. Thetested subjects include 265 and 947 persons underwent fullovernight polysomnography from two different sleepcenters. With 265 subjects from one center as the trainingset and 147 subjects from the other center as the validationset, the first part of our experiments yields SWS detectionresults of Kappa coefficients 0.72-0.78, sensitivity 0.77-0.90and positive predictive value 0.73-0.82. Using the 947subject dataset, the second part of the experimentscompares the relative merits of the tested methods andinvestigates the impacts of SWS ratio and severity of sleepapnea on the performances of the proposed methods.Finally, our results suggest that the quality of the trainingset is of great importance for the development of accurateSWS detection methods.

Index Terms—slow wave sleep, EEG, EOG, automaticsleep staging

Cite: Liang-Wen Hang, Bo-Lin Su, and Chen-Wen Yen, "Detecting Slow Wave Sleep via One or Two Channels of EEG/EOG Signals," International Journal of Signal Processing Systems,  Vol. 1, No. 1, pp. 84-88,  June 2013. doi: 10.12720/ijsps.1.1.84-88