DOI

10.17077/drivingassessment.1670

Location

Santa Fe, New Mexico, USA

Date

25-6-2019

Session

Session 1 – Driver Behavior, Distraction and Crash Risk

Abstract

The goal of this study is to design a novel framework incorporating deep-learning techniques and wearable sensors to recognize manual distractions during driving. Manual distraction is defined as hands off the wheel for any reason (e.g. trying to get a cell phone). In this preliminary study, participants were tasked to drive in city street and highway scenarios in a driving simulator. Verbal instructions prompted participants to perform various manual distraction tasks. The motion of driver’s right wrist during driving was recorded by a wearable inertial measurement unit. A deep-learning technique called convolutional neural network (CNN) was then constructed and trained based on 72% of the experiment trials, and evaluated by the remaining 28% of trials. The results indicated that the convolutional neural network is able to recognize the type of manual distraction task based on the right wrist motion with 87.0% accuracy and F1-score of 0.87. The results indicated that there is a good potential to apply deep-learning techniques and wearable sensing to monitor driver’s inattention status.

Rights

Copyright © 2019 the author(s)

DC Citation

Proceedings of the Tenth International Driving Symposium on Human Factors in Driver Assessment, Training and Vehicle Design, 24-27 June 2019, Santa Fe, New Mexico. Iowa City, IA: Public Policy Center, of Iowa, 2019: 22-28.

Share

COinS
 
Jun 25th, 12:00 AM

Recognition of Manual Driving Distraction Through Deep-Learning and Wearable Sensing

Santa Fe, New Mexico, USA

The goal of this study is to design a novel framework incorporating deep-learning techniques and wearable sensors to recognize manual distractions during driving. Manual distraction is defined as hands off the wheel for any reason (e.g. trying to get a cell phone). In this preliminary study, participants were tasked to drive in city street and highway scenarios in a driving simulator. Verbal instructions prompted participants to perform various manual distraction tasks. The motion of driver’s right wrist during driving was recorded by a wearable inertial measurement unit. A deep-learning technique called convolutional neural network (CNN) was then constructed and trained based on 72% of the experiment trials, and evaluated by the remaining 28% of trials. The results indicated that the convolutional neural network is able to recognize the type of manual distraction task based on the right wrist motion with 87.0% accuracy and F1-score of 0.87. The results indicated that there is a good potential to apply deep-learning techniques and wearable sensing to monitor driver’s inattention status.