Driving EEG based multilayer dynamic brain network analysis for steering process. (30th November 2022)

• Record Type: Journal Article
• Authors: Chang, Wenwen; Meng, Weiliang; Yan, Guanghui; Zhang, Bingtao; Luo, Hao; Gao, Rui; Yang, Zhifei
• Is Part Of: Expert systems with applications. Volume 207(2022)
• Publication Date: 2022-11-30
• Subjects: Multi-layer Networks -- Functional Connectivity -- Electroencephalogram (EEG) -- Driving Intention -- Feature Extraction -- Driving Behavior; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2022.118121
• Languages: English
• ISSNs: 0957-4174

Abstract

Highlight: Apply dynamic functional brain network analysis for driving behavior study. Multilayer network shows significant difference between turning and straight driving. Multilayer modularity and participation coefficient can be used for steering action recognition. Proposed a novel framework for the analysis of dynamic characteristic of EEG. Abstract: Objectives: EEG-based brain computer interface has been demonstrated to be an effective tool for brain state and driving behavior detection to understand the human factors during driving. By providing a driving assistance operation consistent with the driver's action intention, it can improve the interaction process between driving system and its driver. Driving is a comprehensive process that requires the coordination of different brain regions. Functional connectivity, especially the dynamic connectivities calculated by statistical interdependencies between neural oscillations within these brain regions, which can provide some specific information for driving behavior. Methods & experiments: We developed a novel multi-layer brain network model for steering action to improve the understanding of dynamic characteristics during driving. Firstly, a simulated driving experiment is designed and participants were required to drive along a specified route to complete the left turn, right turn and straight action when arriving at an intersection, and electroencephalographic (EEG) signals were recorded simultaneously using a 32-channel system. Then, a multi-layer network framework which combined with an oscillatory envelope based functional connectivity metrics was designed to present the dynamic process of the driving. Results: The result shows there exist significant difference in the multi-layer network structure among the three steering conditions, especially between steering and straight moving. The corresponding parameter analysis also found the significant difference of multilayer modularity (Q-value) and multiplex participation coefficient (MPC) value among the three conditions. Further analysis about single network found the averaged degree, global efficiency, and clustering coefficient also shows significant difference between straight moving and steering action. Conclusion: We conclude that the multi-layer network model can more truly present the dynamic process during driving and provide more accurate information from spatial domain. Besides, the MPC and Q-Value are two new network markers can be used for the recognition of expected steering action, while the average value of corresponding super-matrix can also be used for straight driving and steering action recognition. Implication: The results demonstrate the feasibility of multilayer dynamic brain networks in driving behavior recognition, provided a new insight for the EEG based driving behavior recognition.