Maximum Likelihood Based Classification for the Microstructure of Human Sleep Joss Allen Lima 1'2 jalima@isr.ist.utl.pt Agostinho Rosa 1 acrosa@isr.ist.utl.pt 1Institute for Systems and Robotics Instituto Superior T6cnico 2Labelec SA Av Rovisco Pais 1 Torre Norte 6.21 1000 Lisbon Portugal (phone) +351.1.8418277, (fax) +351.1.8418291 Abstract - In this paper a classifier for the microstructure paradigm of human sleep, the Cyclic Alternating Pattern Sequence (CAPS), is presented. Sleep electroencephalogram (EEG) is the signal used for the scoring. An EEG model makes the feature extraction (pre-processing). The CAPS phases are then detected using maximum likelihood (ML) estimation. A final processing block checks CAPS context rules. This system was tested with good results on the record of 8 hours sleep of a normal adult subject. Keywords - Sleep EEG, Microstructure, CAPS, EEG model, Maximum Likelihood estimator I. INTRODUCTION The most common criteria for sleep EEG was presented in 1968 by Rechtschaffen and Kales [8]. It segmented 8 hours of sleep in 7 different stages: Awake (W), Movement Time (MT), Rapid Eye Movement (REM), and four stages of Non REM sleep (1NREM-4NREM). The length of each epoch is between 20 to 30 seconds. By that, this scoring criteria is also known as a
/lp/association-for-computing-machinery/maximum-likelihood-based-classification-for-the-microstructure-of-kstDVG0ikD