Measuring elemental time and duty cycle using automated video processing.

Area of Science:

• Biomechanics
• Human-Computer Interaction
• Ergonomics

Background:

• Assessing hand kinematics in repetitive tasks is crucial for understanding ergonomics and fatigue.
• Current methods often require markers or are intrusive, limiting their applicability.
• Automated, marker-less solutions are needed for objective and unobtrusive kinematic analysis.

Purpose of the Study:

• To develop and validate marker-less 2D video algorithms for measuring hand kinematics.
• To evaluate the performance of decision tree (DT) and feature vector training (FVT) algorithms in analyzing repetitive tasks.
• To assess the accuracy of these algorithms in determining hand activity levels (HAL) and duty cycle (DC).

Main Methods:

• A marker-less 2D video algorithm was employed to capture hand location, velocity, and acceleration during paced repetitive laboratory tasks.
• The decision tree (DT) algorithm identified hand trajectories using spatiotemporal relationships.
• The feature vector training (FVT) method utilized a k-nearest neighbour classifier trained on task samples.

Main Results:

• The DT algorithm achieved an average duty cycle (DC) error of 2.7%.
• The FVT algorithm demonstrated an average error of 3.3% when trained on the first cycle and 2.8% when trained on representative cycles.
• Errors for hand activity levels (HAL) were negligible (0.1) for both algorithms, with elemental time measurements not statistically different from ground truth (p < 0.05).

Conclusions:

• Both marker-less algorithms (DT and FVT) effectively and automatically measure elapsed time, DC, and HAL in repetitive tasks.
• The developed approach is automatic, repeatable, objective, and unobtrusive.
• This technology is suitable for evaluating repetitive exertions, muscle fatigue, and manual tasks in various settings.