Mental Fatigue Detection of Construction Equipment Operators Based on EEG Collected by a Novel Valve-Type Semidry Electrode Using Deep Residual Shrinkage Networks Under Real Construction Environment

• Published in: IEEE transactions on instrumentation and measurement Vol. 73; pp. 1 - 14
• Main Authors: Wang, Fuwang, Chen, Daping, Zhang, Xiaolei
• Format: Journal Article
• Language: English
• Published: New York IEEE 2024; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accuracy; Algorithms; Classification algorithms; Construction equipment; Construction equipment operators; Data acquisition; deep residual shrinkage network (DRSN); Electrodes; Electroencephalography; electroencephalography (EEG); Fatigue; Feature recognition; Impedance; mental fatigue; Noise measurement; novel valve-type semidry electrode; Operators (personnel); Signal classification; Signal quality; Skin; Springs
• ISSN: 0018-9456; 1557-9662
• DOI: 10.1109/TIM.2024.3451583

Dry electrodes used for electroencephalography (EEG) signals acquisition are of good portability. However, the large impedance of the dry electrode often results in poor signal acquisition quality. Meanwhile, the preparation of wet electrodes in experiments is complex, and replenishing the conductive gel is not convenient. Therefore, based on the comprehensive consideration of the trade-off between wet and dry electrodes used for data acquisition, a novel valve-type semidry electrode is proposed. The semidry electrode has the advantages of convenient wearing and convenient replenishment of conductive liquid. In addition, this experiment is conducted in a real building construction environment, and the acquired EEG signals are highly susceptible to being mixed with various noises. In this study, a deep residual shrinkage networks (DRSNs) algorithm with strong antinoise ability is used to construct a recognition model of construction equipment operators' mental fatigue using EEG-based features. The DRSN algorithm removes mixed noise and redundant features to improve the recognition accuracy of mental fatigue. These results show that the novel valve-type semidry electrode can collect EEG signals with fatigue characteristics just as well as the wet electrode. The ability to maintain small impedances for a long time can benefit long-period EEG signal acquisition. Additionally, the DRSN algorithm was used to classify the mental fatigue state of 12 subjects. The average classification accuracy is 99.59%; the average precision is 96.40%; the average recall is 97.45%; and the average F1 score is 96.93%. The DRSN algorithm outperforms existing approaches in terms of classification accuracy and noise resistance. This study developed the novel valve-type semidry electrode and the DRSN algorithm to successfully measure mental fatigue changes in construction equipment operators over an extended period in a real building construction setting.