Finding Robust Low Dimensional Features for Sleep Detection Using EEG Data

Thumbnail Image

Files

MaoWidjajaGuo22_LowDimensionalFeaturesForSleepDetection_Final.pdf (526.95 KB)

Links to Files

https://ieeexplore.ieee.org/abstract/document/9988155

Permanent Link

https://doi.org/10.1109/ICDSCA56264.2022.9988155
 http://hdl.handle.net/11603/28156

Collections

UMBC Computer Science and Electrical Engineering Department
UMBC Faculty Collection

Author/Creator

Author/Creator ORCID

https://orcid.org/0000-0003-1650-5524

Date

2022-12-29

Type of Work

conference papers and proceedings
postprints
Text

Department

Program

Citation of Original Publication

H. X. Mao, J. Widjaja, Y. Guo, J. Yin and R. Vinjamuri, "Finding Robust Low Dimensional Features for Sleep Detection Using EEG Data," 2022 IEEE 2nd International Conference on Data Science and Computer Application (ICDSCA), Dalian, China, 2022, pp. 42-45, doi: 10.1109/ICDSCA56264.2022.9988155.

Rights

© 2022 IEEE.  Personal use of this material is permitted.  Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

Subjects

Abstract

The study of sleep is crucial for understanding how our bodies function., and electroencephalogram (EEG) offers a convenient way to examine sleep. Sleep can be categorized into wakefulness., rapid eye movement (REM) sleep., and stages 1-4, with 4 being the deepest stage of sleep. We strive to study how well EEG data help classify brain waves into these stages. The goal of this paper is to construct low dimensional features that are computationally efficient, robust., and effective for sleep detection and analysis. We experiment with EEG band power analysis., principal component analysis (PCA)., and autoencoder to reduce the dimensionality of the EEG data and evaluate their performances in classification. We find that, even when highly compressed., two dimensional features are still sufficient to obtain satisfactory classification accuracies: 89.3%, 88.8%, and 90% from band power analysis (using delta and alpha waves), PCA., and autoencoder, respectively.