Eye Glance and Head Turn Correspondence during Secondary Task Performance in Simulator Driving

Abstract

Driver eye glance behavior is a useful input to estimate driver visual attention to the road scene. However, eye tracking in an automobile is currently both expensive and prone to various types of degradation. Because head rotation provides a larger target for sensing than eyeball rotation, head pose monitoring may sometimes serve as a surrogate for driver eye glances. To investigate the correspondence between eye gaze and head pose while driving on a simulated rural highway, participants were asked to take 1, 2 or 4 glances to read aloud single capital letters presented on a scrolling LED display placed at various locations: Left Mirror (LM), Left Cluster (LC), Lower Lap (LL), Center Mirror (CM), Center Stack (CS), and Center Console (CC). Manual eye glance and head pose data reduction from video revealed a relatively low likelihood (65% or less) of head turns when reading from the cluster or center mirror. On the other hand, there was a high likelihood (93% or more) of one or more head turns when reading from the lap (e.g., a smartphone held low), left mirror, center console, and center stack locations, regardless of requested number of glances. This study verifies that there may be opportunities for head pose detection and algorithm development to be pursued in future research.