An Identification Method for Road Hypnosis Based on XGBoost-HMM

Longfei Chen, Chenyang Jiao, Bin Wang, Xiaoyuan Wang, Jingheng Wang, Han Zhang, Junyan Han, Cheng Shen, Kai Feng, Quanzheng Wang, Yi LiuSensors 2025 · DOI

TL;DR

This paper introduces a novel method for identifying "road hypnosis," a dangerous state of reduced awareness in drivers, by fusing EEG, eye movement, and vehicle data. Their XGBoost-HMM model achieves high accuracy (94.9%) in detecting this state, offering a significant advancement for vehicle active safety systems.

What Problem Does This Paper Address?

Human factors are a leading cause of traffic accidents, with distracted and fatigued driving being major contributors. "Road hypnosis" is a specific, dangerous state where drivers become unconsciously disengaged from their surroundings due to prolonged, monotonous driving, leading to delayed reactions and increased crash risk. Existing research often focuses on fixed, monotonous routes. This paper aims to address the challenge of identifying road hypnosis in more realistic, non-fixed driving scenarios by integrating multiple data sources and employing advanced machine learning techniques.

Methods

The authors propose a road hypnosis identification model that combines the strengths of the Hidden Markov Model (HMM) and XGBoost.

  • Data Fusion: They collect data from three modalities:
  • EEG: Electroencephalography data to capture brain activity.
  • Eye Movement: Data from eye-tracking sensors to monitor gaze, pupil diameter, and fixation.
  • Vehicle Data: Speed and acceleration collected via a mobile phone app.
  • Feature Extraction:
    • EEG: Power Spectral Density (PSD) analysis is used to extract features from different frequency bands (delta, theta, alpha, beta, gamma).
  • Eye Movement: Local Linear Embedding (LLE) is employed to extract features from preprocessed eye movement data, preserving local geometric structures.
  • Vehicle Data: Sliding window analysis and point-by-point calculations are used to extract features like average speed, speed fluctuation, average acceleration, and acceleration rate of change.
  • Model Architecture:
    • HMM: Used to model the temporal dynamics and transitions between latent driver states (normal vs. road hypnosis).
  • XGBoost: Integrates with HMM to optimize the observation probability model, allowing it to handle complex, nonlinear relationships within the multimodal data and address class imbalance.
  • Interpretability: LIME and SHAP methods are used to understand the model's decision-making process and the contribution of individual features.
  • Validation: K-fold cross-validation is used to assess the model's generalization ability and stability.

Datasets

The study utilizes data collected from 30 participants during real-world driving experiments on their fixed, non-monotonous commuting routes. This dataset, featuring EEG, eye movement, and vehicle data, appears to be a newly collected benchmark for road hypnosis research in varied driving environments.

Key Insights and Ideas

  • Multimodal Fusion is Crucial: Combining EEG, eye movement, and vehicle data significantly improves road hypnosis identification accuracy compared to single modalities.
  • EEG is Highly Informative: Beta and delta waves from EEG consistently emerge as the most critical features for identifying road hypnosis, highlighting the importance of brain activity monitoring.
  • Speed and Pupil Diameter are Key Indicators: Vehicle speed and right pupil diameter also show strong importance across various metrics, suggesting their significant role in detecting this state.
  • Non-Monotonous Routes are Relevant: Road hypnosis is not limited to highways; it can occur in complex urban driving, making the identification of this state in diverse scenarios vital.
  • XGBoost Enhances HMM: Integrating XGBoost with HMM effectively handles the complex, multimodal data and improves the model's predictive power.

Key Results

  • High Accuracy: The proposed XGBoost-HMM model achieved an accuracy of 94.9% in identifying road hypnosis states.
  • Superior Performance: This model outperformed traditional HMM and Random Forest (RF) models, which achieved accuracies of 92.66% and 92.66% respectively.
  • Feature Importance: EEG features (beta, delta waves) were found to be the most important, followed by pupil diameter right and vehicle speed. Acceleration was identified as a less significant feature.
  • Computational Efficiency: XGBoost-HMM demonstrated better computational efficiency than Random Forest in terms of training time, inference time, memory usage, and model size.

Limitations and Future Work

  • Small Sample Size: The study acknowledges that a key limitation is the small sample size (30 participants), which may affect the generalizability of the findings. Future work should consider larger sample sizes to improve robustness.
  • Controlled Conditions: Experiments were conducted under clear weather and good road conditions. Future research should simulate diverse weather conditions (rain, snow, smog), road conditions (slippery, bumpy), and time changes (night driving) using virtual driving environments.
  • Further Validation: While the model shows strong performance, further validation in more diverse and challenging real-world driving scenarios would be beneficial.

Summary updated

Abstract

Human factors are the most important factor in road traffic crashes. Human-caused traffic crashes can be reduced through the active safety system of vehicles. Road hypnosis is an unconscious driving state caused by the combination of external environmental factors and the driver's psychological state. When drivers fall into a state of road hypnosis, they cannot clearly perceive the surrounding environment and make various reactions in time to complete the driving task, and driving safety is greatly affected. Therefore, road hypnosis identification is of great significance for the active safety of vehicles. A road hypnosis identification model based on XGBoost-Hidden Markov is proposed in this study. Driver data and vehicle data related to road hypnosis are collected through the design and conduct of vehicle driving experiments. Driver data, including eye movement data and EEG data, are collected with eye movement sensors and EEG sensors. A mobile phone with AutoNavi navigation is used as an on-board sensor to collect vehicle speed, acceleration, and other information. Power spectrum density analysis, the sliding window method, and the point-by-point calculation method are used to extract the dynamic characteristics of road hypnosis, respectively. Through normalization and standardization, the key features of the three types of data are integrated into unified feature vectors. Based on XGBoost and the Hidden Markov algorithm, a road hypnotic identification model is constructed. The model is verified and evaluated through visual analysis. The results show that the road hypnosis state can be effectively identified by the model. The extraction of road hypnosis-related features is realized in non-fixed driving routes in this study. A new research idea for road hypnosis and a technical scheme reference for the development of intelligent driving assistance systems are provided, and the life identification ability of the vehicle intelligent cockpit is also improved. It is of great significance for the active safety of vehicles.
BibTeX
@inproceedings{chen2025identification,
  author = {Longfei Chen and Chenyang Jiao and Bin Wang and Xiaoyuan Wang and Jingheng Wang and Han Zhang and Junyan Han and Cheng Shen and Kai Feng and Quanzheng Wang and Yi Liu},
  title = {{An Identification Method for Road Hypnosis Based on XGBoost-HMM}},
  booktitle = {Sensors},
  year = {2025},
  doi = {10.3390/s25061842},
}
Figure