Detecting Driver Distraction

Source Organization:T-SET UTC
University:Carnegie Mellon University
Principal Investigator:Maxine Eskenazi and Alan Black
PI Contact Information:6413 Gates Hillman Complex 412-268-3858
Project Manager:
Funding Source(s) and Amounts Provided (by each agency or organization):$50,000 - US Dept of Transport
Total Dollars:$50,000
Agency ID/Contract/Grant Number:
Start and End Dates:
Project Status:Complete
Subject Categories:Safety and human factors
Abstract:It is well known that texting during driving causes divided attention, higher cognitive load and accidents. At the same time, many people feel that hands-free talking on the phone is not dangerous. Yet the same type of effort (understanding a conversation, figuring out what to say, feeling emotions, etc) is present when speaking. While it might be less objectionable at lower speeds on limited access roads, as the amount of issues to deal with increases and as speed increases, driving and speaking on the phone are a very dangerous combination. While there has been much work on detecting distraction in the driver’s habits, there is much less on detecting distraction from the point of view of their speech patterns. In order to find a way to alert drivers to imminent danger, we will develop algorithms that automatically detect, from the driver’s speech, when they are too distracted to be driving. We will first collect data on a simulator under several conditions: changes in speed, changes in unexpected events on the road, changes in amount of information in the conversation. We will mine this data to find indicators that change as conditions deteriorate along these three parameters. Later work will use the data collected to develop an automatic distraction detection system.
Describe Implementation of Research Outcomes (or why not implemented):
Impacts/Benefits of Implementation (actual, not anticipated):
Project URL: