Artificial Intelligence designed for analysing human activity in a work environment

Abstract

The rise of work-related musculoskeletal disorders (WMSDs) has become a major concern in various industries, leading to serious health problems and economic losses. Despite the automation of some manufacturing processes, manual tasks are still necessary and can pose ergonomic risks to workers. To address this issue, an AI-powered tool for ergonomic risk assessment has been developed. The tool successfully estimates the 3D human pose with a mean per joint position error (MPJPE) of 46.8 mm, using the Human3.6M dataset, and calculates the Rapid Entire Body Assessment (REBA) score in real time, providing a comprehensive assessment of ergonomic risk factors. Our approach has been validated by a specialist doctor in rehabilitation. The system employs a semi-supervised learning approach with a fully convolutional model based on dilated temporal convolution over 2D keypoints. The developed AI-powered tool provides immediate feedback, enabling enhanced actions for risk reduction. Case studies demonstrate the effectiveness of the approach for improving the accuracy and efficiency of ergonomic risk assessment in various industries.