An EEG Based Vehicle Driving Safety System Using Automotive CAN Protocol

SBCE; V.Mathew, Renji; Basheer, Jasmin · 2015 · Crossref

DOI: 10.14445/22315381/ijett-v26p236

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 drowsy driving, which contributes significantly to traffic accidents, injuries, and fatalities. Existing drowsiness detection systems primarily rely on monitoring physical behaviors such as eye blinking, head nodding, or gripping force. However, these methods are prone to false alarms because physical signs do not always correlate with mental fatigue, particularly in cases where drivers experience "mental sleep" with eyes open. To overcome these limitations, the authors propose a real-time drowsiness detection system that analyzes brain activity using electroencephalography (EEG). The system utilizes a single-channel, dry-electrode EEG sensor (NeuroSky Mindwave) to capture brain signals wirelessly, offering a more comfortable and practical alternative to traditional wet-electrode systems that require skin preparation and conductive gel. The proposed system architecture consists of a brain-computer interface, a data processing unit, and vehicle nodes implemented on ARM7 cores (LPC2148). Communication between these nodes is facilitated via the automotive Controller Area Network (CAN) protocol. The operational workflow begins with the driver wearing the sensor and blinking three times to initiate the ignition, ensuring the sensor is active. The system continuously monitors the driver’s attention level. If attention drops below a specific threshold, an audible alert is triggered. If drowsiness persists, the system commands the Engine Control Unit (ECU) to activate hazard signals and eventually switch the vehicle into a self-controlled deceleration mode, guided by proximity sensors, until the driver regains alertness. The core detection algorithm processes raw EEG data by applying a Type 1 Chebyshev band-pass filter (0.5–50 Hz) to remove artifacts, followed by a Fast Fourier Transform (FFT) to convert signals into the frequency domain. Power spectra are calculated using 512-point moving windows and converted to a logarithmic scale. A Support Vector Machine (SVM) classifier is then employed to distinguish between alert and drowsy states. The study tested four kernel functions: linear, polynomial, radial basis function (RBF), and sigmoid. Results indicated that while delta and theta power remained relatively stable, alpha and beta power increased during drowsiness. Among the kernels tested, the RBF kernel yielded the highest accuracy, achieving 85% successful classification (17 out of 20 trials) with a 15% error rate, outperforming linear (70%), polynomial (75%), and sigmoid (65%) kernels. The significance of this work lies in its demonstration of a feasible, non-invasive method for real-time driver monitoring that directly addresses mental fatigue rather than just physical symptoms. By integrating EEG analysis with automotive CAN protocols, the system provides a robust mechanism for preventing accidents caused by drowsiness. The authors conclude that while the prototype successfully achieves up to 85% accuracy, future implementations may require manufacturer assistance for seamless vehicle integration. The study highlights the potential for expanding such systems to allow for more advanced mind-controlled vehicle operations, emphasizing the reliability and efficiency of CAN bus technology in automotive networks.

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-nvidia on 2026-08-10; verification: verified.

Topics

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

• drowsiness detection algorithms
• drowsiness

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
• Theoretical Contribution: computational model