Extraction of parameters for lane change intention based on driver's gaze transfer characteristics. (June 2020)

• Record Type: Journal Article
• Authors: Peng, J.S.; Wang, C.W.; Fu, R.; Yuan, W.
• Is Part Of: Safety science. Volume 126(2020)
• Publication Date: 2020-06
• Subjects: Lane change -- Driving experiment -- Intention representation parameters -- Gaze transition -- Markov theory -- Road safety; Industrial accidents -- Periodicals; Accident Prevention -- Periodicals; Safety -- Periodicals; Travail -- Accidents -- Périodiques; 363.11
• DOI: 10.1016/j.ssci.2020.104647
• Languages: English
• ISSNs: 0925-7535

Abstract

Highlights: Fixation time of lane-change intention period is lower than in lane-keeping period. Saccade time/average saccade velocity is higher in lane-change intention period. Eye and head movement reflects a strong correlation in lane-change intention period. One/Two-step gaze transition show visual search process/paths in different periods. There are more main visual scanpaths related to the right in lane-keeping period. Abstract: In naturalistic driving environment, lane change is a necessary, frequent operation, and causes traffic accidents and fatalities. Current lane-change driver assistance system uses turn signals as start time; however, drivers show lane-change intention before turning on the signals. To identify lane-change intention, a driving experiment was designed and conducted, during which eye movement characteristics were recorded from 16 drivers. Time period data from 5 s before changing lanes and 5 s in lane-keeping period were analysed. Saccade speed, eye–head coordination mode, and other parameters were confirmed to be significantly different between two periods. Based on area division and Markov theory, attributes related to gaze transition were obtained, including one- and two-step gaze transition probabilities, main gaze transition paths, and regional gaze probabilities. In summary, all features extracted from eye movement characteristics, gaze transition paths, and gaze areas demonstrate that visual attention is more dispersed during the lane-change intention period. These parameters also lay the foundation for lane-change intention recognition systems and make lane change safer.