Brain Signal Classification Using Self-tuning Assisted Fuzzy Structure Uncertain Indirect Observer


TayebiHaghighi S., Lee Y., Koo I.

4th International Conference on Intelligent and Fuzzy Systems (INFUS), Bornova, Turkey, 19 - 21 July 2022, vol.504, pp.794-801 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Volume: 504
  • Doi Number: 10.1007/978-3-031-09173-5_91
  • City: Bornova
  • Country: Turkey
  • Page Numbers: pp.794-801
  • Keywords: Brain signal classification, Sleep stage classification, Self-tuning approach, Fuzzy technique, Structure uncertain observer
  • Yıldız Technical University Affiliated: No

Abstract

Sleep staging is a critical step that can help to identify sleep disturbances. Recently, sleep stages classification is accomplished a serious issue in preserving people's lives. Tiredness and drowsiness in driving can endanger the lives of many people in vehicle accidents. Thus, identifying sleep disorders, which include the identification and classification of various sleep stages is an important subject. Biomedical signals such as an electroencephalogram (EEG) are used to recognize sleep disorders. In this research, a self-tuning indirect estimation approach along with a nonlinear modeling technique is proposed for brain signal classification with high accuracy. This proposed approach consists of five steps. In the first stage, the brain signals are resampled, and the root means square (RMS) feature is extracted from resampled brain signals. After that, in the second step, the resampled RMS brain signals are modeled using Gaussian autoregressive-Laguerre approach. To improve the accuracy and robustness, in the next step, the proposed self-tuning fuzzy technique along with structure uncertain observer is recommended. In the fourth step, the RMS resampled residual brain signal is generated. Based on the difference in the levels of the RMS resampled residual brain signals and based on support vector machine (SVM), the brain signals will be classified into alertness, ambiguous, drowsiness, and sleep modes. According to the results, the classification accuracy using the proposed method is around 98%.