A single-channel EEG-based drowsiness detection system using IoT technology and a reconfigurable set of five power spectral density features achieved 95% accuracy in detecting driver drowsiness while overcoming interpersonal variability due to aging.
Key Findings
Results
The proposed EEG-based drowsiness detection system achieved an accuracy of 95% using a reduced set of features and a single differential EEG channel.
The system used a single EEG channel combined with IoT technology for cloud-based alert signaling.
Five features were computed from the power spectral density (PSD).
The system was validated using the MIT-BIH Polysomnography dataset comprising ten subjects.
The approach targeted detection of the transition from wakefulness to drowsiness (stage one of sleep).
Results
The system is person-dependent and designed to overcome interpersonal variability due to aging.
The system employs a reconfigurable set of features to address individual differences between subjects.
Interpersonal variability was identified as a key challenge in EEG-based drowsiness detection.
High detection accuracy was maintained despite this variability.
The person-dependent design was described as the main advantage of the proposed system.
Methods
Feature extraction was based on variations in power spectral energy during the transition from wakefulness to drowsiness.
A set of five features was computed from the power spectral density.
Features captured changes in power spectral energy specifically during the wakefulness-to-drowsiness transition.
The feature set was described as 'reconfigurable,' allowing adaptation across individuals.
The reduced feature set was sufficient to achieve 95% accuracy.
Background
The system incorporated IoT technology to send alert signals to the cloud.
Cloud connectivity was integrated into the embedded system design.
Alert signals are transmitted remotely, enabling real-time drowsiness monitoring.
The system was described as a 'connected embedded system,' combining signal processing with network communication.
Methods
The MIT-BIH Polysomnography dataset with ten subjects was used to validate the system.
The dataset comprised ten subjects.
Polysomnography data provided multi-channel physiological recordings from which EEG data were used.
The dataset is a publicly available benchmark used for sleep-stage classification and drowsiness studies.
What This Means
This research suggests that it is possible to reliably detect driver drowsiness using just a single EEG (brainwave) sensor combined with internet-connected technology. The system analyzes changes in brainwave energy patterns as a person transitions from being fully awake to becoming drowsy, using five specific measurements derived from those patterns. By tailoring the system to each individual person rather than using a one-size-fits-all approach, the researchers were able to achieve 95% accuracy in detecting drowsiness across ten test subjects, even accounting for natural differences between people due to age and other factors.
A key practical feature of this system is that it can send real-time alerts to the cloud when drowsiness is detected, which could allow for remote monitoring or automated warnings in vehicles. Using only a single EEG channel rather than many electrodes simplifies the hardware considerably, making the system more practical and wearable for real-world use.
This research matters because drowsy driving is a major cause of traffic accidents worldwide, and current detection technologies are often too complex, expensive, or inaccurate for widespread use. A simple, accurate, and internet-connected drowsiness detector like the one described here could represent a meaningful step toward practical safety systems for drivers, though further validation with larger and more diverse populations would be needed before real-world deployment.
Belakhdhar I. (2026). Connected Embedded System for Drowsiness Detection Based on a Reconfigurable Set of Features.. Sensors (Basel, Switzerland). https://doi.org/10.3390/s26041195