Monte Carlo-based Measurement Uncertainty Propagation on a Virtual Sensor for Real- Time Human Balance Exergame Analysis

Nowadays, increasingly advanced Artificial Intelligence techniques are progressively being integrated with design methodologies. The use of advanced inferential models trained on real-time data is making it possible to analyze the dynamic behaviors of very complex systems simply by observing them through vision-based measurement techniques, offering an innovative approach to indirect measurement of forces at play without the need for additional bulky and difficult-to-install sensors. However, it is not always possible to estimate the reliability of these systems because they depend on variables that are poorly controlled and difficult to assess. In addition, the assessment of measurement uncertainty is critical for the adoption of data-driven artificial intelligence models in commercial products and services. This paper explores the possibility of using cameras for the development of a virtual sensor for the dynamic analysis of human balance in a balance platform system. The developed virtual sensor is characterized by estimating the propagation of the measurement uncertainty through the use of the Monte Carlo method. The results obtained make it possible to evaluate the metrological performance of the proposed method.