Scholarlink

Drowsy driving emerges as a major factor contributing to traffic accidents. Drowsiness is characterized by a feeling of tiredness and a compelling desire to sleep. It is evident through a gradual decrease in Reaction Time (RT) of the driver. Reaction Time (RT) refers to the duration taken for a person or system to respond to a given sudden unexpected stimuli or event. Accurate prediction of Reaction Time (RT) to unexpected events is crucial for enhancing safety and performance. Electroencephalogram (EEG), capturing the brain’s electrical signals, demonstrates the most substantial correlation with drowsiness. Consequently, EEG is broadly recognized as a trustworthy tool for assessing drowsiness, fatigue, and overall performance. While many studies have utilized traditional machine learning and deep learning techniques to detect drowsiness or alertness from EEG data, there has been limited research focused on accurately predicting Reaction Time (RT). In this study, we aim to forecast drivers’ reaction times using EEG data through a novel Covariance 2D CNN-LSTM framework. Here, the objective is to forecast Reaction Time (RT) based on a 5-second EEG trial that precedes it. The superiority of the proposed method was validated through regression metrics, specifically Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE), and was compared against current state-of-the-art methods.

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

EEG-Based Reaction Time Prediction Using Covariance Augmented 2D Convolutional Neural Network

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Adarsh V. Parekkattil,
  • Sanjeev Kumar Varun,
  • Tharun Kumar Reddy Bollu

摘要

Drowsy driving emerges as a major factor contributing to traffic accidents. Drowsiness is characterized by a feeling of tiredness and a compelling desire to sleep. It is evident through a gradual decrease in Reaction Time (RT) of the driver. Reaction Time (RT) refers to the duration taken for a person or system to respond to a given sudden unexpected stimuli or event. Accurate prediction of Reaction Time (RT) to unexpected events is crucial for enhancing safety and performance. Electroencephalogram (EEG), capturing the brain’s electrical signals, demonstrates the most substantial correlation with drowsiness. Consequently, EEG is broadly recognized as a trustworthy tool for assessing drowsiness, fatigue, and overall performance. While many studies have utilized traditional machine learning and deep learning techniques to detect drowsiness or alertness from EEG data, there has been limited research focused on accurately predicting Reaction Time (RT). In this study, we aim to forecast drivers’ reaction times using EEG data through a novel Covariance 2D CNN-LSTM framework. Here, the objective is to forecast Reaction Time (RT) based on a 5-second EEG trial that precedes it. The superiority of the proposed method was validated through regression metrics, specifically Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE), and was compared against current state-of-the-art methods.