Measuring Cyclists' Behavior Using Wearable Sensors for Automatic Creation of Safety Behavior Map

According to the traffic accident report by the National Police Agency, more than 150,000 persons are injured in bicycle-related traffic accidents in 2009 in Japan. In order to reduce bicycle-related traffic accidents, collecting cyclists' actual behavior data especially safety-related behavior data under various traffic conditions is essential. In this paper, we propose (1) a method for continuously measuring and analyzing cyclists' turning-head behavior (including both safety visual search behavior and look-away behavior) using wearable sensors, (2) an automated method to identify whether the measured turning-head behavior is safety visual search behavior or look-away behavior requiring no prior information, and (3) a method for automatic creation of cyclists' safety behavior map showing traffic spots where many cyclists perform safety visual search behavior. Our method use small wireless three-axis gyro sensors to measure cyclists' head motion. Since gyro sensors and a PDA are both small, wireless and running on battery power, our method can easily adapt to various kinds of situations. By applying independent component analysis (ICA) to reduce bicycle-caused noise, our method allows to detect cyclists' turning-head behavior with precision of 90.5% and recall of 91.8%. In addition, by focusing on spatial density of turning-head behavior among cyclists, identification ratio of visual search behavior achieved 18% improvement without using any prior information. Furthermore, safety behavior map created from 36 junior high school students’ data showed the fact that almost 60% of all subjects never perform visual search behavior even when entering blind intersections including accident prone area.

• Corporate Authors:

  Transportation Research Board

  500 Fifth Street, NW
  Washington, DC  United States  20001
• Authors:
  • Tada, Masahiro
  • Noma, Haruo
  • Renge, Kazumi
• Conference:
  • Transportation Research Board 90th Annual Meeting
  • Location: Washington DC, United States
  • Date: 2011-1-23 to 2011-1-27
• Date: 2011

Language

• English

Media Info

• Media Type: DVD
• Features: Figures; Photos; References; Tables;
• Pagination: 14p
• Monograph Title: TRB 90th Annual Meeting Compendium of Papers DVD

Subject/Index Terms

• TRT Terms: Automatic data collection systems; Behavior; Bicycle crashes; Cost effectiveness; Cyclists; High risk locations; Safety
• Geographic Terms: Japan
• Subject Areas: Data and Information Technology; Pedestrians and Bicyclists; Safety and Human Factors; I80: Accident Studies;

Filing Info

• Accession Number: 01337230
• Record Type: Publication
• Report/Paper Numbers: 11-3233
• Files: TRIS, TRB
• Created Date: Apr 19 2011 7:17AM