Driver Drowsiness Classification Using Fuzzy Wavelet-Packet-Based Feature-Extraction Algorithm

Khushaba, R N; Kodagoda, S; Lal, S; Dissanayake, G · 2010 · OpenAlex-citations

DOI: 10.1109/tbme.2010.2077291

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the critical safety issue of driver drowsiness, a significant contributor to motor vehicle accidents characterized by impaired awareness and reduced vigilance. The authors aim to develop an automated method for classifying driver drowsiness levels using physiological signals, specifically Electroencephalogram (EEG), Electrooculogram (EOG), and Electrocardiogram (ECG). The research is motivated by limitations in existing detection methods, which often rely on Fast Fourier Transform (FFT) techniques unsuitable for nonstationary physiological signals, require a large number of sensor channels, or use entropy measures like Shannon entropy that are better suited for compression than classification. To overcome these issues, the study proposes a novel feature extraction algorithm called Fuzzy Mutual Information based Wavelet Packet Transform (FMIWPT). The methodology employs the Wavelet Packet Transform (WPT) to decompose physiological signals into tree-structured subspaces, allowing for simultaneous time and frequency localization. Unlike previous methods that use fixed spectral bands, this approach explores the full spectrum to identify features that best discriminate between alert and drowsy states. The core innovation is the use of fuzzy mutual information to evaluate the discriminatory power of these features. By utilizing fuzzy memberships to estimate entropy and mutual information, the method reduces computational costs associated with traditional density estimators while providing accurate information content measures. The algorithm automatically selects the optimal wavelet packet decomposition for each subject, thereby accounting for individual variability in EEG dynamics. The experimental design involved thirty-one male volunteer drivers aged 20–69 who performed simulated driving tasks. Data was collected using only three EEG channels, one EOG channel, and one ECG channel to ensure practicality. Participants underwent an initial alert driving session followed by a monotonous driving session designed to induce drowsiness, with physiological signals recorded continuously. The results demonstrate the efficacy of the FMIWPT method in extracting features that highly correlate with different drowsiness levels. The proposed algorithm achieved an average classification accuracy of 95%–97% across all subjects. This performance highlights the method's ability to handle the nonstationary nature of biosignals and its robustness against individual variability, which often compromises group-statistic-based approaches. The use of fuzzy mutual information proved superior to traditional entropy measures in identifying the most relevant frequency components for drowsiness detection. Furthermore, the study confirms that high classification accuracy can be maintained with a minimal number of sensor channels, addressing a key limitation of prior research that required extensive channel setups. The significance of this work lies in its contribution to the development of practical, real-time driver drowsiness detection systems. By reducing the number of required sensors and improving classification accuracy through advanced feature extraction, the FMIWPT method offers a more feasible solution for integration into vehicle safety systems. The findings underscore the importance of using adaptive, subject-specific feature selection methods rather than fixed spectral bands or global statistics. This approach not only enhances the reliability of drowsiness detection but also provides a framework for processing other nonstationary biomedical signals where individual variability and signal complexity are prevalent.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified_with_issues.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• drowsiness detection algorithms
• drowsiness
• microsleep

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: physiological data
• Methodological Resource: tool software, validation psychometrics