ADTIDO: Detecting the Tired Deck Officer with Fusion Feature Methods

Area of Science:

• Maritime Safety
• Biomedical Engineering
• Signal Processing

Background:

• Maritime accidents can be reduced by monitoring deck officer fatigue.
• Electroencephalogram (EEG)-based systems offer fatigue detection but face individual variability and signal sensitivity issues.
• Video-based methods analyzing eye characteristics are an alternative, but integrating them with EEG can improve accuracy.

Purpose of the Study:

• To introduce the Automatic Detect the TIred Deck Officers (ADTIDO) algorithm for classifying deck officer fatigue levels.
• To enhance EEG-based fatigue detection by integrating video-based eye closure data.
• To develop a novel definition of fatigue based on individual real-time operational reaction time.

Main Methods:

• Utilized Discrete Wavelet Transformer (DWT) to decompose EEG signals into six sub-signals for feature extraction (MAV, SD, RMS).
• Developed a video-based approach to obtain Eyelid Closure Degree (ECD) values from EEG signals, differing from traditional video methods.
• Created ECD-EEG fusion features by combining ECD and EEG feature sets for classifier input.

Main Results:

• The Bidirectional Gated Recurrent Unit (Bi-GRU) network achieved 90.19% classification accuracy using only EEG features, outperforming other classifiers.
• Integrating ADTIDO's fused ECD-EEG features resulted in a final classification accuracy of 95.74% for deck officer fatigue levels.
• The study validated the algorithm's efficacy through trials with 21 subjects using collected EEG signals.

Conclusions:

• The ADTIDO algorithm effectively combines EEG and video-based eye closure data for accurate deck officer fatigue detection.
• Fusion of ECD-EEG features significantly improves fatigue classification performance compared to using EEG features alone.
• This approach offers a promising method for enhancing maritime safety by proactively identifying and mitigating officer fatigue.