ANN based Joint Time and frequency analysis of EEG for detection of driver drowsiness

Dabbu, Suman; Malini, M.; Reddy, B. Ram; Vyza, Yashwanth Sai Reddy · 2017 · Crossref

DOI: 10.14429/dlsj.2.10370

archive: archived pipeline: cataloged

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

Summary

This study addresses the critical need for a reliable quantitative index to detect driver drowsiness, a major contributor to road traffic accidents. While physiological signals like EEG are known to reflect brain states, previous methods lacked a standardized metric for quantifying drowsiness levels. The research aims to bridge this gap by proposing a Drowsiness Index (DI) derived from joint time and frequency analysis of EEG signals, validated against subjective assessments and classified using an Artificial Neural Network (ANN). The experimental design involved thirteen male volunteers with over two years of driving experience who performed a 60-minute monotonous driving task on a static simulator. Eight-channel EEG data was acquired at 1000 Hz from frontal, parietal, occipital, and temporal lobes. The signals were pre-processed using Chebyshev filters and decomposed into Delta, Theta, Alpha, and Beta rhythms. Feature extraction focused on two key parameters: Power within Root Mean Square Deviation (PRMSD) in the time domain and Power within Spectrogram (PSG) in the frequency domain, calculated using Short-Time Fourier Transform. These features were statistically analyzed using Friedman tests and linear regression, and subsequently used to train a back-propagation ANN classifier with a 2-2-1 architecture (two inputs, two hidden neurons, one output). The results demonstrated that PRMSD and PSG values significantly decreased as drowsiness increased, with a statistically significant correlation ($\rho < 0.05$) between the total mean and drowsy mean of subjects ($R^2$ ranging from 0.82 to 0.96). The study established a clear separable marker for drowsiness: active states corresponded to index values of 0.65–1.0, drowsy states to 0.01–0.4, and a transition zone of 0.4–0.6. The ANN classifier achieved high performance, with a training accuracy of 99.81%, testing accuracy of 99.80%, sensitivity of 99.82%, and specificity of 99.78%. The derived drowsiness index showed complete agreement with ratings provided by three neuro-physicians, confirming its validity as a quantitative measure of alertness. This work is significant for developing advanced driver assistance systems by providing a robust, data-driven method to quantify drowsiness in real-time. The high accuracy of the ANN-based classification suggests that EEG-derived features can effectively distinguish between active and drowsy states, offering a viable pathway for automated accident avoidance technologies. The study confirms that beta and alpha rhythm power reductions are reliable indicators of fatigue, supporting the integration of physiological monitoring into vehicle safety protocols.

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 qwen3.8-27b-gittensor on 2026-08-23; verification: pending re-verification.

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