Non-contact Sensing for VDT Worker State Assessment （Part 1）

Evaluation of Accuracy in Peak Detection for rPPG

Abstract

The assessment of VDT （Visual Display Terminal） workers’ states is essential for their health management and optimizing the work environment. Autonomic nervous system indices, EEG, and eye movement-related indices are useful for assessing arousal levels and attention; however, conventional measurement methods require sensor or device attachment, which poses a burden on workers and leads to measurement deficiencies caused by work movements. Therefore, we developed a non-contact measurement method using a general-purpose RGB camera to simultaneously capture multiple indices. This report focuses on pulse wave analysis and implements an SVM-based method to detect falsely detected peaks for removal and completion. Data were collected and analyzed from 13 healthy adult males, who provided informed consent, during approximately 8 minutes of VDT work. The accuracy of pulse wave peak detection was evaluated by using peaks obtained from fingertip photoplethysmography as the ground truth, yielding a false positive rate of 5.85％ and a false negative rate of 5.61％. After applying the proposed method, the comparison of the mean pulse rate achieved an MAE of 0.70 bpm and an RMSE of 0.44 bpm, demonstrating performance suitable for practical use. These results indicate that the proposed method is useful for estimating the mean pulse rate without burdening VDT workers.