AbstractConstruction workers’ poor mental states can lead to numerous safety and productivity issues. One major trend in construction research is quantitatively evaluating workers’ psychophysiological states. With the advances in wearable electroencephalogram (EEG) devices, such assessment can be possible by interpreting workers’ brainwave patterns. However, the recorded EEG signals are highly contaminated with signal noises, particularly ocular-related artifacts generated from blinking and eye movement. Although most of the noise can be suppressed by well-established filtering techniques, ocular artifacts cannot be eliminated easily and automatically by conventional techniques. To overcome this challenge, this study proposes a procedure to reduce ocular artifacts by integrating dependence component analysis, image processing, and machine learning algorithms. The results demonstrated the potential of the proposed procedure to produce high-quality EEG signals accurately, continuously, and automatically during construction operations. The findings contribute to the body of knowledge by overcoming the barriers to reliable translation of EEG signals in numerous construction-related investigations, especially those that add substantially to the understanding of the effect of workplace stressors on workers’ health and safety.