借助chat GPT阅读文献——基于改进的时间-图卷积网络的端到端EEG信号疲劳驾驶分类模型_end-to-end fatigue driving eeg signal detection mo

文章主要内容

• 目前研究存在的问题：Most recent methods only consider the features of individual electrode EEG signals, ignoring the functional connectivity of the brain.

• 本文提出的解决办法：Proposing a MATCN-GT model. Using GCN to process EEG signal. Graph data can be used to represent the functional connectivity of the brain. GCN is good at learning the internal structure information of EEG signals.

• 达到的效果：The accuracy of the MATCN-GT model reached 93.67%, outperforming existing algorithms.

文献原文

End-to-end fatigue driving EEG signal detection model based on improved temporal-graph convolution network

ABSTRACT

Fatigue driving is one of the leading causes of traffic accidents, so fatigue driving detection technology plays a crucial role in road safety. The physiological information-based fatigue detection methods have the advantage of objectivity and accuracy. Among many physiological signals, EEG signals are considered to be the most direct and promising ones. Most traditional methods are challenging to train and do not meet real-time requirements. To this end, we propose an end-to-end temporal and graph convolution-based (MATCN-GT) fatigue driving detection algorithm. The MATCN-GT model consists of a multi-scale attentional temporal convolutional neural network block (MATCN block) and a graph convolutional-Transformer block (GT block). Among them, the MATCN block extracts features directly from the original EEG signal without a priori information, and the
GT block processes the features of EEG signals between different electrodes. In addition, we design a multiscale attention module to ensure that valuable information on electrode correlations will not be lost. We add a Transformer module to the graph convolutional network, which can better capture the dependencies between long-distance electrodes. We conduct experiments on the public dataset SEED-VIG, and the accuracy of the MATCN-GT model reached 93.67%, outperforming existing algorithms. Furthermore, compared with the traditional graph convolutional neural network, the GT block has improved the accuracy rate by 3.25%. The accuracy of the MATCN block on different subjects is higher than the existing feature extraction methods.

1. Introduction

Drivers driving for a long time or driving at night will cause their physical and psychological functions to decline, affecting their normal driving. Fatigue driving affects the drivers’ driving skills, mental state, concentration, thinking and judgment, and also senses and perceptions. In serious cases can lead to reduced motor ability, which can easily lead to traffic accidents. In 2004, the World Health Organization released the ‘‘World Report on Road Traffic Injury Prevention’’, which pointed out that fatigue driving is one of the important risk factors causing traffic accidents. According to the forecast of the World Health Organization, the number of road traffic deaths will rise to approximately 2.4 million per year by 2030, and road traffic deaths will rise to become the fifth leading cause of death worldwide [1]. As the number of casualties due to fatigue driving continues to increase, fatigue driving detection technology has become increasingly important. The existing fatigue detection methods mainly include physiological information-based, vehicle information-based, and facial feature-based. Among them, physiological information-based detection methods can directly reflect the drivers’ driving status. Among the many physiological signals, the electroencephalographic (EEG) signals are considered the most direct and promising.

The driver facial feature-based detection method infers the driver’s fatigue state through eye state, mouth state and head posture. The method mainly uses a camera to capture the driver’s facial images, and extracts fatigue-related information through computer vision technology. Azhar et al. [2] used long and short-term memory (LSTM) to capture eye-related movements. The method uses 1-D LSTM (RLSTM) as the basis and convolutional LSTM (C-LSTM) to process the image, through two LSTMs to simulate eye movement. Huang et al. [3] proposed a multi-granularity deep convolutional model (RF-DCM) for fatigue driving detection. He designed a multi-granularity extraction subnetwork to overcome the problem of head pose variation, proposed a feature recalibration subnetwork to learn local features, and finally fused global and local features with a feature fusion network. The driver facial feature-based detection method is susceptible to factors such as lighting and occlusion, decreasing accuracy and reliability.

The vehicle information-based detection method indirectly judges the driver’s fatigue state according to the driver’s vehicle handling. The method uses on-board sensors and cameras to collect data parameters such as steering wheel steering angle and grip, vehicle speed, and vehicle driving route, and analyzes the differences in driving behavior parameters between normal driving and fatigue conditions to determine the driver’s fatigue status. Li et al. [4] found that the difference between the driver’s maximum/minimum grip on the steering wheel and the standard grip is closely related to the driver’s fatigue state, so they designed a fatigue driving detection method based on the driver’s grip on the steering wheel. Lu et al. [5] detected driver fatigue by embedding surface electromyography (sEMG) sensors on the steering wheel. This method collects the driver’s physiological signals non-invasive, which can detect early driver fatigue. The vehicle information-based detection method can make it challenging to collect accurate and stable data due to the driver’s driving habits and different
proficiency levels.

EEG signal is the spontaneous or evoked electrical activity of nerve cell groups in specific parts of the brain during physiological processes, which reflects the biological activity of the brain and contains a large amount of information. From the perspective of electrophysiology, every subtle activity of the human brain triggers the discharge of the corresponding nerve cells, which special instruments can record to analyze and decode brain activity. EEG decoding is the separation of task-relevant components from the EEG signals. The main method of decoding is to describe the task-related components through feature vectors, and then use classification algorithms to classify the relevant feature vectors for different tasks. The decoding accuracy depends on how well the features extracted by the feature algorithm represent the relevant tasks, and how accurately the classification algorithm can separate the different tasks. The EEG signal records the electrical wave changes during brain activity, which is the most direct and effective reflection of fatigue. EEG waves are classified into four types according to their amplitude and frequency: