Multimodal Detection of Driver Distraction

Using driving scenarios and a driving simulator, where the driver interacts with a voice-driven email app that systematically controls cognitive load, the project team has collected 50 sessions (50 different drivers, about 25 minutes each), containing user distraction.  The project team is formatting this database to make it publicly available. After the demo at the University Transportation Center's (UTC’)s recent DC safety session, the team has seen interest in obtaining the database from transportation colleagues. Last year the project team identified two specific phenomena that they found in distracted driver's speech: fillers (ums, and ers), and hesitations (the amount of silence). The team built classifiers that automatically detect a speaker’s distraction based on those aspects of their speech (and without the linguistic content, the words they said).  These detectors have been recently integrated into a Yahoo email reader platform (the Yahoo InMind agent) as an android app. The detection has been demonstrated to Yahoo and at conferences. It will be integrated in the Yahoo news reader app within the next two months. This use of the detector informs the driver by gracefully shutting down an app when distraction has been detected and restarting where it left off when it is safe to do so.

    Language

    • English

    Project

    • Status: Completed
    • Funding: $127,500
    • Sponsor Organizations:

      Technologies for Safe and Efficient Transportation University Transportation Center

      Carnegie Mellon University
      Pittsburgh, PA  United States  15213

      Office of the Assistant Secretary for Research and Technology

      University Transportation Centers Program
      Department of Transportation
      Washington, DC  United States  20590
    • Managing Organizations:

      Carnegie Mellon University

      Pittsburgh, PA  United States 
    • Project Managers:

      Ehrlichman, Courtney

    • Performing Organizations:

      Carnegie Mellon University

      Pittsburgh, PA  United States 
    • Principal Investigators:

      Eskenazi, Maxine

    • Start Date: 20170101
    • Expected Completion Date: 20180730
    • Actual Completion Date: 0
    • USDOT Program: University Transportation Centers Program
    • Subprogram: Distracted driving

    Subject/Index Terms

    Filing Info

    • Accession Number: 01645842
    • Record Type: Research project
    • Source Agency: Technologies for Safe and Efficient Transportation University Transportation Center
    • Files: UTC, RiP
    • Created Date: Sep 6 2017 7:01PM