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.

L1-norm based time-varying brain neural network and its application to dynamic analysis for motor imagery

Time-Varying Effective Connectivity for Describing the Dynamic Brain Networks of Post-stroke Rehabilitation

Dynamical Differential Covariance Based Brain Network for Motor Intent Recognition

The Dynamic Brain Networks of Motor Imagery: Time-Varying Causality Analysis of Scalp EEG

The Time-Varying Network Patterns in Motor Imagery Revealed by Adaptive Directed Transfer Function Analysis for Fmri.

Exploiting Dimensionality Reduction and Neural Network Techniques for the Development of Expert Brain-Computer Interfaces

Tensor-based Dynamic Brain Functional Network for Motor Imagery Classification

Short time Fourier transformation and deep neural networks for motor imagery brain computer interface recognition.

Constructing Time-Varying Directed EEG Network by Multivariate Nonparametric Dynamical Granger Causality

Dynamic Modeling of Common Brain Neural Activity in Motor Imagery Tasks

A Dynamic Domain Adaptation Deep Learning Network for EEG-based Motor Imagery Classification

Improving Generalization Based on L1-Norm Regularization for EEG-Based Motor Imagery Classification.

Driving EEG based multilayer dynamic brain network analysis for steering process

Dynamic Joint Domain Adaptation Network for Motor Imagery Classification

A Spatio-Temporal Interactive Attention Network for Motor Imagery EEG Decoding

A framework for motor imagery with LSTM neural network

A novel recognition and classification approach for motor imagery based on spatio-temporal features

EEG-Based Multi-Frequency Multilayer Network for Exploring the Brain State Evolution Underlying Motor Imagery

A Multifrequency Brain Network-Based Deep Learning Framework for Motor Imagery Decoding

Exploring Motor Imagery Eeg Patterns for Stroke Patients with Deep Neural Networks

One-Dimensional Convolutional Multi-branch Fusion Network for EEG-Based Motor Imagery Classification