Fusing Multiple Sensor Modalities for Complex Physiological State Monitoring
Abstract
Both the public sector and the military are working on developing drowsiness detection systems, as driver fatigue is a significant contributor to motor vehicle accidents. Individually, electroencephalography (EEG) and eye-tracking measures are tenuous indicators of driver fatigue and impairment. This project proposes to integrate multiple sensor modalities in order to improve drowsiness level assessment and driver performance prediction. There is substantial evidence supporting the correlation of alpha bursts in EEG (narrowband alpha power density increases lasting 500 ms to several seconds) and eye-tracking measures, such as pupil diameter and gaze distribution, with drowsiness. As a step towards multi-sensory data fusion, we aim to implement in real time an optimized version of an existing algorithm for the automatic detection of alpha bursts using a single EEG channel and ascertain correlations between alpha bursts, eye-tracking measures, and behavioral indicators of fatigue that include standard deviation of both lane position and acceleration. The ability to reliably detect alpha bursts in real-time combined with established correlations will allow an algorithm to accurately predict driver performance in a simulation environment.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 2012
- Accession Number
- ADA568853
Entities
People
- Brent Lance
- Daniel B Silversmith
- Justin Brooks
- Kaleb G. McDowell
- Kesshi M Jordan
- Nicholas Perkons
- Scott Kerick
- Wiliam Hairston
- William Nothwang
Organizations
- United States Army Research Laboratory