Comparison of the observer, single-frame video and computer vision hand activity levels

Area of Science:

• Occupational Health
• Ergonomics
• Computer Vision

Background:

• The Hand Activity Level (HAL) is a crucial metric in occupational health for assessing risk.
• Traditional methods for measuring HAL include observer ratings and manual video analysis.
• Advancements in computer vision offer potential for automated and efficient HAL assessment.

Purpose of the Study:

• To compare the accuracy and consistency of automated computer vision (HAL_C) with observer (HAL_O) and single-frame video analysis (HAL_F) for Hand Activity Level.
• To evaluate the feasibility of using computer vision for industrial hygiene assessments.

Main Methods:

• Analysis of 419 industrial videos from three prospective cohort studies.
• Measurement of Hand Activity Level using three distinct methods: observer rating (HAL_O), single-frame video analysis (HAL_F), and automated computer vision (HAL_C).

Main Results:

• Statistically significant differences were observed between the three HAL measurement methods (p < .001), though average differences were small (<1).
• Computer vision (HAL_C) showed high consistency with single-frame video analysis (HAL_F), with 68% of cases within ±1 HAL scale point.
• A strong linear regression (R² = 0.89) was found between HAL_C and HAL_F, indicating comparable results.

Conclusions:

• Automated computer vision methodology provides results comparable to single-frame video analysis for Hand Activity Level assessment.
• Computer vision presents a viable and consistent alternative for industrial task analysis and occupational risk evaluation.