Effects of interactive modality on the spatiotemporal characteristics of driver eye movement
Open Access
Article
Conference Proceedings
Authors: Lin Jie, Xing Liu, Tingru Zhang, Alan Chan
Abstract: In-vehicle information system (IVIS), as crucial components in the vehicles, provide drivers with convenient functionalities but also pose potential safety hazards. Operating these systems requires visual attention, potentially increasing the risk of accidents. While previous researches focused on static eye metrics like fixation and saccade, limited attention has been given to spatiotemporal eye movement characteristics crucial for information acquisition while driving. This study investigated the impacts of three modalities (voice-based, touchscreen-based, and gesture-based) on spatiotemporal characteristics of driver eye movement. Thirty-six participants were recruited to a simulated driving experiment, with one group acting as baseline without non-driving related tasks (NDRTs), while others performed NDRTs using one of different interactive modalities. Scanpaths, fixation entropy, and visual transition probability matrices were analyzed to understand spatiotemporal characteristics. A new comparison method based on ScanMatch algorithm was proposed to measure the similarity of scanpaths. The K-means clustering was used to identify areas of interest (AOIs), while Shannon’s equation was applied to calculate fixation entropy. Visual transition probability matrices normalized transition counts, revealing areas with the most transitions. Results showed the voice group's eye movements closely resembled the baseline, with higher entropy in driving-related AOIs. In contrast, the touchscreen group had lower entropy and a higher likelihood of distraction. Thus, voice-based interaction had the least distracting effect, resembling baseline eye movement patterns. These findings offer insights for designing safer IVIS interactions to reduce traffic accidents.
Keywords: In-vehicle information systems, Interaction modality, Eye movement patterns
DOI: 10.54941/ahfe1005234
Cite this paper:
Downloads
78
Visits
138