High-speed rail (HSR) accidents are largely attributed to human errors made by train drivers. HSR drivers frequently encounter situations that induce fatigue during operations, leading to a decrease in their vigilance. Therefore, conducting online research on changes in HSR drivers’ vigilance and implementing timely fatigue interventions are crucial to ensure that they can safely and attentively perform their tasks.

This paper proposes a novel method for identifying and addressing driver fatigue by leveraging electroencephalogram (EEG) signal characteristics and develops a wearable alertness intervention device that integrates real-time monitoring, fatigue detection, and intelligent intervention. This system includes a collection terminal, web-based fatigue detection equipment, fatigue intervention modules, and Bluetooth microcontrollers. This study involved collecting real-time fatigue EEG data utilizing an eight-channel EEG apparatus within the Stroop fatigue induction paradigm, which was then transmitted via Wi-Fi to a web-based platform. The OPENBCI open-source software converted EEG waveforms into digital signals. MATLAB software employed the Welch algorithm to extract seven key EEG indicators, including five basic waveform powers, power ratios, and comprehensive feature values. These indicators were analyzed alongside subjective fatigue assessments (KSS values) and objective reaction times to examine variations in driver EEG waves. The grey correlation analysis method was employed to determine the weight of each EEG indicator with respect to objective reaction time. A KSS value of 5 or higher was used as the fatigue benchmark, with a weighted average establishing a reaction time of 1.118s as the criterion for fatigue evaluation. The random forest algorithm, implemented in Python with hyper parameter optimization via grid search, was used for weighted feature extraction of fatigue EEG indicators to establish fatigue threshold values for real-time fatigue recognition. Intervention commands were transmitted to the main control circuit board of the device via LoRa remote communication. This paper designed and developed physical intervention modules for smell, sound, vibration, and electrical stimulation that were controlled and integrated using double-layer PCB circuit boards. These were ultimately integrated with a flexible wearable alertness intervention device that provides diverse and personalized intervention effects. Additionally, we designed an alertness warning platform to visualize the driver’s alertness status and fatigue interventions.

According to experimental data, the EEG-based fatigue detection classifier demonstrated over 90% accuracy on the test set. Each intervention module showed consistent communication performance and effectively boosted alertness. The system maintained stable operation for more than 2.5 hours.

This paper presents the design of a portable, wearable alertness intervention device leveraging online EEG signal detection technology to monitor and intervene in HSR fatigue. It offers a thorough solution to address fatigue-related issues in HRS drivers. The study’s findings provide valuable theoretical and practical guidelines for designing and implementing fatigue intervention devices, enhancing driver safety and performance. Ultimately, these results offer fresh data references for rostering and duty planning, with significant potential for further advancement and use.

Driving safety on high-speed rail (HSR) is of utmost importance, with a direct correlation between HSR travel safety and driver vigilance levels. This work aimed to develop a robust and comprehensive experimental scheme for estimating vigilance in HSR drivers.

The methodology integrates multichannel data to realize a nuanced and accurate evaluation of HSR driver vigilance levels. To simulate and reduce alertness among HSR drivers, a state-of-the-art HSR driving simulator is used, which shows continuous driving scenarios that closely imitate real-world conditions. Within this simulated environment, the basis for a multichannel HSR driver’s vigilance estimation research and experimental scheme is established. The experimental phase involves the simultaneous collection of different data points, including the driver’s EEG, ECG, eye movements, response times to simple tasks, and subjective reports detailing the level of fatigue experienced. These synchronous multichannel datasets, which are rich in information, form the basis for developing a sophisticated driver vigilance estimation model. Machine learning methods, such as neural networks and support vector machines, are exploited to leverage the wealth of multisource fusion features, which include the incorporation of the driver’s neural and physiological characteristics as the input and the driver’s fatigue state as the output. Central to this work is a careful examination of the effectiveness of the developed multichannel vigilance estimation scheme for HSR drivers, verifying its validity and practical application in real-world HSR driving scenarios.

The experimental design highlights the significance of considering multichannel data and identifying the intricate interplay between occupational responsibilities and environmental factors that influence HSR drivers. The comprehensive experimental plan spans the collection of diverse data sources, including EEG, ECG, eye movement recordings, response times to various tasks, and subjectively reported levels of vigilance. The driver’s response time to emergency situations is normalized to establish a vigilance grading standard, such as high, medium, and low, allowing for the identification of the driver’s vigilance. This nuanced classification system offers the foundation for proactive measures to improve driving safety and prevent potential railway accidents or delayed responses during crucial situations. Acknowledging the constraints of wearable devices in collecting and transmitting physiological data, we advocate a forward-looking approach. Follow-up studies should encompass a holistic consideration of hardware and software aspects. This includes addressing the challenges associated with the collection, transmission, processing, and analysis of multichannel data. Proposals for improvements in data collection devices and transmission methods are presented to minimize interference and fortify the overall robustness of the experimental design.

In conclusion, this comprehensive approach establishes a solid theoretical and technical foundation for the preliminary design of safety systems for high-speed train applications. By addressing the challenges associated with multichannel data collection and wearable devices, this work considerably contributes to the advancement of HSR driver vigilance estimation and, consequently, plays a key role in improving the overall driving safety in HSR operations.