Automatic microsleep detection based on KNN classifier utilizing selected and effective EEG channels
Annually, the global economy suffers significant financial losses due to decreased productivity of work, accidents, and crashes in traffic resulting from microsleep. To reduce the adverse impacts of microsleep, it is necessary to have a discreet, dependable, and socially acceptable method of detecti...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Universiti Teknologi Malaysia
2024
|
| Subjects: | |
| Online Access: | http://umpir.ump.edu.my/id/eprint/43898/ http://umpir.ump.edu.my/id/eprint/43898/1/Automatic%20microsleep%20detection%20based%20on%20KNN%20classifier.pdf |
| Summary: | Annually, the global economy suffers significant financial losses due to decreased productivity of work, accidents, and crashes in traffic resulting from microsleep. To reduce the adverse impacts of microsleep, it is necessary to have a discreet, dependable, and socially acceptable method of detecting microsleep episodes consistently throughout the day, every single day. Regrettably, the current solutions fail to match these specified criteria. Moreover, by utilizing sophisticated features and employing machine learning techniques, it is possible to process electroencephalogram (EEG) information in a highly efficient manner, enabling the rapid and successful detection of microsleep. The selection of an optimum channel and the use of a competent classification algorithm are crucial for effective microsleep detection. One unique channel selecting strategy has been introduced in the current study to evaluate the classifying accuracy of microsleep detection based on EEG. This strategy is based on correlation coefficients and utilizes the K-Nearest Neighbor (KNN) method. Furthermore, the Fast Fourier Transform (FFT) was employed for extracting the feature, so validating the endurance of the proposed technique. In order to enhance the speed of the microsleep detecting system, the study was performed using 3 distinct time windows: 0.5s, 0.75s, and 1s. The study revealed that the suggested approach achieved a classification accuracy of 98.28% within a time window of 0.5 seconds to detect microsleep using EEG signal. The exceptional effectiveness of the given system can be efficiently utilized in detecting microsleep using EEG signal. |
|---|