Scholarlink

<p>Driver drowsiness remains a significant safety concern on the road. This study explores the potential of probabilistic machine learning techniques to enhance driver drowsiness detection and classification. By leveraging EEG and heart rate data, we develop a novel framework that achieves a remarkable 100% accuracy in identifying driver wakefulness states. Our approach in-volves preprocessing EEG data, extracting power spectral densities, and modeling their relationship with heart rate using Support Vector Regression (SVR). Subsequently, Bayesian Support Vector Classification (SVC) is employed to categorize drivers as awake, drowsy, or sleepy. While the model demonstrates exceptional performance, the predicted class probabilities reveal the in-fluence of additional factors beyond EEG and heart rate. Despite the limitations of relying solely on EEG and heart rate data, our findings highlight the potential of probabilistic machine learning for accurate driver drowsiness detection. Future research could explore the integration of other physiological signals, such as EOG, EMG, skin conductance, and breath rate, to further improve model reliability.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Unveiling driver drowsiness: a probabilistic machine learning approach using EEG and heart rate data

  • 期刊论文

  • 发表日期：

    2025年6月10日

查看全文
  • Abolfazl Khishdari,
  • Hamid Mirzahossein

摘要

Driver drowsiness remains a significant safety concern on the road. This study explores the potential of probabilistic machine learning techniques to enhance driver drowsiness detection and classification. By leveraging EEG and heart rate data, we develop a novel framework that achieves a remarkable 100% accuracy in identifying driver wakefulness states. Our approach in-volves preprocessing EEG data, extracting power spectral densities, and modeling their relationship with heart rate using Support Vector Regression (SVR). Subsequently, Bayesian Support Vector Classification (SVC) is employed to categorize drivers as awake, drowsy, or sleepy. While the model demonstrates exceptional performance, the predicted class probabilities reveal the in-fluence of additional factors beyond EEG and heart rate. Despite the limitations of relying solely on EEG and heart rate data, our findings highlight the potential of probabilistic machine learning for accurate driver drowsiness detection. Future research could explore the integration of other physiological signals, such as EOG, EMG, skin conductance, and breath rate, to further improve model reliability.