IoT Based AI Application for Posture Recognition to Reduce Workplace Injuries

Details

• Personal Author:
  Deshpande AM ; Kumar M ; Kumar R
• Description:
  Motivation: Lifting heavy items is one of the leading causes of injury in the workplace. As per U.S. Bureau of Labor Statistics report: In 2001, over 36 percent of injuries involving missed workdays resulted due to shoulder and back injuries in workers. In 2017, 97,990 workers taking a day away from work involving overexertion in lifting or lowering. Some of the common reasons for workplace injuries: Short term pains in due to accidental events such as worker getting struck by some object. Overexertion during manual heavy object transportation. The weight lifting pains can result in: Inflammation and trauma to the tissue in and around the joints. Wear and tear of the joints and cartilage. Degeneration of the tendons. Early-onset of arthritis. Posture detection and monitoring can loop to serve as an early warning system for the workers to enhance their work efficiency and physical health. The motivation behind the proposed study is to address the problem of detecting and monitoring body postures in the work place where heavy object lifting is a part of the worker routine. Background: Real-time pose estimation for a person is a key component in enabling machines to have an understanding of people in images and videos. Papandreou et al. (2017), Cao et al.(2017), Cao et al. (2018) have proposed various approaches for human pose recognition using deep learning. We intend to develop a human pose monitoring system on top of these approaches. We also intend to further predict the behavioral intent of the person using temporal data and recurrent models of neural network in our system. Being aware of good posture is the first step to breaking old poor postural habits and reducing stress and strain on your spine. Objectives and Task Description: The main objective of this project is detecting and monitoring body postures in the work place where heavy object lifting is a routine. Specific tasks involved in this project are as follows: Design and development of an Edge Computing based camera feed processing unit to enable real-time posture detection on factory floors. Develop user interface to allow multi-device accessibility and data visualization. System design to allow prediction of workers' behaviors using computer vision and deep learning. Experimental Design: We are using the postural cuttoffs defined in REBA (Rapid Entire Body Assessment) standards as a reference for our development. The data-sets of human activity understanding presented by Caba Heilbron et al. (2015) and human pose dataset presented by Andriluka et al. (2014) will be used for this project. No human subjects will be involved in the current phase. Expected Results: An IoT device using NVIDIA Jetson TX2 board as the processing unit. The setup will be able to suggest corrective measures based on the detected postures of people. Body posture monitoring will be valuable to formulate requirements for a healthy working environment. Optimization of the manual handling of heavy objects on the factory floor is the primary focus of this research. Limitations and Future Direction: Limitations of this system include: This approach being non-invasive, it can't give the physical information about the environment. Including some wearable devices in addition to video feed for posture monitoring has a potential to improve the accuracy of the system. For wide area coverage at workplace, multiple cameras will be required. Future work towards improvement of this system may include: Incorporating the developed application on cloud platform for accessibility on mobile devices. Data collection and data analytic platform development for the human pose recording and intent analysis. Incorporation of developed pose recognition system in robots may result in: Development of robust human robot interaction systems allowing robots to take decisions based on predicted human behavior. This can further ensure safety in case of factory workers working along side manufacturing robots. This module can also help in ensuring safety of pedestrians if used in self-driving cars. [Description provided by NIOSH]
• Subjects:
  Musculoskeletal Diseases Musculoskeletal System Occupational Health Safety Software
• Keywords:
  Artificial Intelligence Computer Models Injuries Injury Prevention Lifting Monitoring Systems MSD Musculoskeletal Disorders Posture
• Publisher:
  University of Cincinnati
• Document Type:
  Text
• Funding:
  Grant
• Genre:
  Abstract or Summary
• Place as Subject:
  Ohio ; OSHA Region 5
• CIO:
  National Institute for Occupational Safety and Health (NIOSH)
• Topic:
  Workplace Safety & Health
• Location:
  North America
• Pages in Document:
  1
• NIOSHTIC Number:
  nn:20058633
• Citation:
  20th Annual Pilot Research Project Symposium, University of Cincinnati Education and Research Center, October 10-11, 2019, Cincinnati, Ohio. Cincinnati, OH: University of Cincinnati, 2019 Oct; :1
• Email:
  deshpaad@mail.uc.edu
• Federal Fiscal Year:
  2020
• Performing Organization:
  University of Cincinnati
• Peer Reviewed:
  True
• Start Date:
  20050701
• Source Full Name:
  20th Annual Pilot Research Project Symposium, University of Cincinnati Education and Research Center, October 10-11, 2019, Cincinnati, Ohio
• End Date:
  20260630
• Collection(s):
  National Institute for Occupational Safety and Health
• Main Document Checksum:
  urn:sha-512:b3154ac2222de85d1562ca738a0a20fe53c677a947e9a7d9128999423823e612777b7889176ad30ff0b49a2d69ea4ec93355ab28636b69e91cf7aac7be83b0c8
• Download URL:
  https://stacks.cdc.gov/view/cdc/221675/cdc_221675_DS1.pdf
• File Type:
  [PDF - 2.15 MB ]