Related Experiment Video
Updated: May 5, 2026
06:37
Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
Published on: December 15, 2023
4.5K
Detection of driver drowsiness using wavelet analysis of heart rate variability and a support vector machine
1Department of Electronic Engineering, Pukyong National University, Busan 608-737, Korea. wychung@pknu.ac.kr.
Sensors (Basel, Switzerland)
|December 10, 2013
Summary
Detecting driver drowsiness using wavelet transform on heart rate variability (HRV) signals significantly improves accuracy. This non-stationary signal analysis method offers a more reliable approach to preventing fatigue-related car accidents compared to traditional methods.
Area of Science:
- Biomedical Engineering
- Transportation Safety
- Signal Processing
Background:
- Fatigued driving poses risks comparable to drunk driving, leading to accidents.
- Current driver drowsiness detection using heart rate variability (HRV) analysis is limited by treating HRV as stationary.
- The wavelet transform is suitable for analyzing non-stationary signals like HRV during driving.
Purpose of the Study:
- To classify alert and drowsy driving states using wavelet transform of short-term HRV signals.
- To compare the performance of wavelet transform-based HRV analysis against traditional Fast Fourier Transform (FFT)-based methods for drowsiness detection.
Main Methods:
- Wavelet decomposition applied to 1-min, 2-min, and 3-min HRV samples.
- Receiver Operating Characteristic (ROC) analysis for feature selection.
- Support Vector Machine (SVM) classifier for distinguishing between alert and drowsy states.
Main Results:
- Wavelet-based HRV analysis demonstrated superior performance over FFT-based methods across all tested durations.
- SVM classification using wavelet features achieved 95% accuracy, 95% sensitivity, and 95% specificity on 1-min HRV signals.
- FFT-based features yielded significantly lower performance: 68.8% accuracy, 62.5% sensitivity, and 75% specificity.
Conclusions:
- Wavelet transform analysis of short-term HRV signals is a highly effective method for real-time driver drowsiness detection.
- The proposed approach offers a substantial improvement in reliability and accuracy over conventional FFT-based techniques.
- An inexpensive and user-friendly hardware platform is feasible for implementing this advanced drowsiness detection system.
