Support vector machine and simple recurrent network based automatic sleep stage classification of fuzzy kernel

  • Recently, sleep disorder is taken as a serious issue in people living. Normally people cerebrum passes through variety of static physiological steps or changes for the duration of sleep. Biomedical signal such as EEG, ECG, EOG and EMG setup and signals used to recognize sleep disorders. This work proposes better technique that can be designed to discriminate the stages of sleep which can help physicians to do an analysis and examination of related sleep disorders. In order to identify a modification inside brain, EEG signal partitioned with 5 frequency bands: delta, theta, alpha, beta and gamma. After signal acquisition, Band pass filter is applied to discriminate the input EEG signal of Fpz–Cz electrodes into frequency bands. Statistical specific features are extracted from distinctiveness impression of EEG signal. Then classification is required for classifying the sleep stages automatically with fuzzy kernel support vector machine and simple recurrent network (SRN). In SRN, statistical features were extracted and allocate 30 s period to 5 possible levels in sleep; wakefulness, Non Rapid Eye Movement Sleep Stage 1 (NREMSS 1), NREMSS 2, NREMSS 3 and NREMSS 4, Rapid Eye Movement Sleep Stage (REMSS). These signal acquired from sleep-EDF repository from PhysioBank (PB) used to validate our proposed scheme. Simple recurrent network classification performance rate is found as 90.2% than that of other new classifiers such as feed forward neural network (FNN) and probabilistic neural network (PNN) next it was compared and results are experimented in proposed work.

  • B.Saravana Balaji
  • Journal of Ambient Intelligence and Humanized Computing
  • 16/06/2020