REM sleep behavior disorder (RBD) without current signs of PD, or any other diseases, is designated as idiopathic RBD (iRBD), and is most likely one of the earliest signs of PD. In long-term prospective studies the percentage of subjects with iRBD, who will eventually develop PD or AP, ranges from 40% to 65% after average 10–15 years. Correct detection of iRBD is therefore essential, especially if treatment of PD becomes available. Our hypotheses is that high EMG activity can be considered as an outlier detection problem, in which atonia, as seen in the healthy elderly, are assumed to be inliers, while abnormally high EMG activity during REM sleep, as seen in iRBD, are considered outliers. A total of forty-eight subjects from the Danish Center for Sleep Medicine, Department of Clinical Neurophysiology, Glostrup University Hospital, Denmark, were enrolled in this study. They were divided into four equal-sized age- matched groups according to their diagnosis; elderly healthy controls, PLMD, iRBD and PD patinets. All involved subjects underwent one full night polysomnography in accordance with the international standard from AASM. The submentalis and the left and right anterior tibialis were analyzed. The recorded EMG signals were pre-filtered to reduce the influence of artifacts. The EMG activity was described by two envelope curves, a baseline-envelope (5 s) and an activity-envelope curve (0.5 s). In this study the envelope curves are obtained by smooting the full- wave-rectified preprocessed EMG signals. The EMG activity was computed as the ratio of the activity-envelope with the minimum of the baseline-envelope. The one-class support vector machine is an unsupervised learning algorithm, which finds the smallest possible boundary that encloses the inliers. The voting principple was used to classify each REM sleep epoch into normal or abnormal. If six or more mini-epochs in each epoch was classified as an outlier, then the whole epoch was labeled as abnormal. The number of abnormal epochs of all REM epochs in percentage was used muscle activity score. It was possible to separate all controls and PLMD patients from iRBD patients. Only 25% of the PD patients were classified as not having RBD. Detection of abnormal high muscle activity during REM sleep can be considered as an outlier detection problem and that iRBD muscle activity is more grouped and complex compared to PLMD activity. We would like to express our gratitude to all those who helped collecting the data, especially Marielle Zoetmulder and Rune Frandsen.