Research Papers

Drowsiness Detection With Electrooculography Signal Using a System Dynamics Approach

[+] Author and Article Information
Dongmei Chen, Zheren Ma, Zeyu Yan, Wei Li

Department of Mechanical Engineering,
University of Texas at Austin,
Austin, TX 78712

Brandon C. Li

The Wharton School of Business,
University of Pennsylvania,
Philadelphia, PA 19104

Contributed by the Dynamic Systems Division of ASME for publication in the JOURNAL OF DYNAMIC SYSTEMS, MEASUREMENT, AND CONTROL. Manuscript received July 25, 2016; final manuscript received December 19, 2016; published online May 15, 2017. Assoc. Editor: Evangelos Papadopoulos.

J. Dyn. Sys., Meas., Control 139(8), 081003 (May 15, 2017) (7 pages) Paper No: DS-16-1368; doi: 10.1115/1.4035611 History: Received July 25, 2016; Revised December 19, 2016

The electrooculography (EOG) signal is considered most suitable for drowsiness detection. Besides its simplicity and low cost, EOG signals are not affected by environmental factors such as light intensity and driver movement. However, existing EOG-based drowsiness detection techniques employ arbitrarily chosen features for classifier training, leading to results that are less robust against changes in the measurement method, noise level, and individual subject variability. In this study, we propose a system dynamics-based approach to drowsiness detection. The EOG signal is treated as a neurophysiological response of the oculomotor system. Each blink action is considered as a result of a series of neuron firing impulses entering the system. Blink signatures are thus extracted to identify the system transfer function, from which system poles are computed to characterize the drowsiness state of the subject. It was found that the location of system poles on the pole–zero map for blink signatures from an alert state was distinctly different from those from a drowsy state. A simple criterion was subsequently developed for drowsiness detection by counting the ratio of real and complex poles of the system over any given period of time. The proposed methodology is a systematic approach and does not require extensive classifier training. It is robust against variations in the subject condition, sensor placement, noise level, and blink rate.

Copyright © 2017 by ASME
Your Session has timed out. Please sign back in to continue.



Grahic Jump Location
Fig. 1

(a) Experimental setup and (b) signal processing procedures

Grahic Jump Location
Fig. 2

An example of a pole–zero map showing two complex conjugate poles, one real pole, and one real zero. The system is a third-order system.

Grahic Jump Location
Fig. 3

Raw EOG signal (a) and filtered EOG signal (b) from record 1. Signal started right before subject went to bed.

Grahic Jump Location
Fig. 4

Filtered EOG signal between minute 4 and minute 7, with peak detection results shown with circles

Grahic Jump Location
Fig. 5

Typical alert (a) and drowsy (b) signatures

Grahic Jump Location
Fig. 6

System identification results: (a) measured and model-predicted alert signal, (b) measured and model-predicted drowsy signal, (c) impulse response of the system model from the alert signal, and (d) impulse response of the system model from the drowsy signal

Grahic Jump Location
Fig. 7

Pole–zero maps of the oculomotor system at (a) alert state and (b) drowsy state

Grahic Jump Location
Fig. 8

Number of complex and real poles of the identified oculomotor system for the first EOG record. The poles were counted every 10 s.

Grahic Jump Location
Fig. 9

System pole counts for the other two EOG records: (a) went to bed alert and fell asleep at about minute 23 and (b) drowsy going to bed and fell asleep at about minute 30




Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related eBook Content
Topic Collections

Sorry! You do not have access to this content. For assistance or to subscribe, please contact us:

  • TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
  • EMAIL: asmedigitalcollection@asme.org
Sign In