Industry
TCS Pace
TCS Pace™: Speedier Digital Transformation & Innovation in Business
Highlights
- Work-related musculoskeletal disorders (WMSDs) that impact the neck, back and wrist are often ignored until they become too painful to continue working.
- There is an identifiable need for an assessment system to forewarn workers or their managers of “high-risk” postures at work.
- TCS proposes a cautionary system that uses computer-based vision to calculate a worker’s posture risk score using real-time or recorded video footage. The assessment that uses machine learning and posture analytics can be sent out as an alert on a wearable or via a centralized device.
- The assessment negates the need for on-site human monitoring by an expert. It uses an industry acceptable assessment system, and plots this on an individual’s body in real-time to pull out posture analytics for a specific individual.
On this page
Work-related injuries are commonplace
Bad posture, whether in a sedentary job or in a task that calls for intense manual labor and movement, puts an employee at serious risk of developing work-related musculoskeletal disorders (WMSD) that tend to be ignored at first, and get worse over time.
As is the norm, employees tend to ignore such symptoms until they become too painful to continue working. This, despite such conditions falling under organizational health, safety, and environment (HSE) rules. A major reason, however, for WMSD being ignored or overseen is the complexity involved in its assessment. There is a clearly identified need to provide workers and managers with evidence of incorrect postures for them to understand, and more importantly, open their eyes to the onset of workplace injury to subsequently improve movement / conditions at the workplace, and consequently maintain productivity as well.
A study on 116 construction workers indicated 32 workers i.e. 27.6% suffered from CTS. Other studies on symptoms of musculoskeletal disorder showed construction workers being the fourth highest in experiencing hand/wrist discomfort, while a survey on 308 apprentice-level electricians and journeymen showed that 144 (47%) individuals frequently experienced problems related to hand/wrist – a disorder that comes second after back-related discomfort experienced by 157 (51%) individuals.
TCS’ use of posture analytics
TCS has proposed a computer vision-based automated method to measure and notify employers about WMSD in real time, using the example of carpal tunnel syndrome (CTS) in the construction industry. CTS is a condition caused by repetitive hand movements, holding on to objects firmly for long periods of time, and other mechanical stressors on the palm that extend to the wrist / forehand.
The program applies a top-down approach where a human body is identified for assessment via an image or through live or pre-existing video footage. Following this, “pose” landmarks are detected in the bounding box area of the body. The program detects a universe of postures to generate what is called a Rapid Entire Body Assessment (REBA) score and indicates warning alerts in red if a particular posture is found to be risky and injury prone. MediaPipe in the python package is used to identify human pose key points in real-time. (MediaPipe is an open-source framework to build pipelines to perform computer vision inference over arbitrary sensory data such as video or audio.)
How machine learning works in the assessment
The package uses machine learning (ML) models that take body pose landmarks in image co-ordinates as an output. The pose land marker model tracks 33 body landmark locations and the angle between two joints is then calculated using a coordinate system. It is important to note that REBA is an assessment method that was formerly done manually. However, TCS Rapid Labs, an innovation team within TCS Pace™, has incorporated ML models of the python REBA package with modifications to calculate body angles when an individual bends using their back or puts a strain on cervical spine. Bending beyond a certain threshold can be identified, indicated in red, and sent out as an alert to the worker or the manager via a wearable or using a centralized monitor or alert system either in real-time or retrospectively.
Here’s what a REBA reading looks like. The threshold values are the ones universally accepted as part of a REBA assessment.
| 0 | Negligible risk. |
| 2-3 | Low risk. Change likely needed. |
| 4-7 | Medium risk. Investigate. Change needed soon. |
| 8-10 | High risk. Investigate. Implement change. |
| 11+ | Very high risk. Implement change. |
The assessment and program can be customized for an employee based on their work environment / industry, and the analysis can be replicated for a range of body parts / postures as well to determine safe-unsafe thresholds. Furthermore, the program can create real-time alerts whenever a REBA threshold is breached, and it can do this using recorded video or a live feed. A representative video below shows an individual in a work position, while a corresponding REBA assessment is drawn in parallel.
A smart option across industries
A computer vision-based automated assessment helps in putting together a slew of measures on worker safety and well-being at the workplace. Firstly, the frequency of assessing WMSDs increases with automated methods.
Secondly, the cost of WMSD assessments can be reduced as it does not require the presence of expert industry specialists since the universal REBA scoring is already an assessment incorporated in the scoring. While an automated REBA scoring approach requires a one-time certification, manual REBA assessment mandates the need for experts to be present.
Furthermore, in the automated approach, CCTV footage otherwise used primarily for security reasons can be repurposed to calculate REBA scores–a smart use of existing resources.
As this approach provides a measurable scientific evaluation and risk quantification method, it makes it possible to calculate an accurate ‘before and after’ impact on workers apropos any employee safety. Interventions as simple as altering how certain tools are used to distribute bodily stress better across joints / tendons, or revisiting work processes to reduce the force and repetitive physical nature of certain tasks, can be incorporated on an immediate basis.
With the advent of superior vision algorithms and developments in chip architectures, individual focus on CTS is possible to identify each joint on the hands. The approach to reduce CTS could be extrapolated to other disorders and injuries under WMSD, to avoid other occupational, health, and safety (OHS) issues. The approach can be extended to industries such as manufacturing and retail, too, apart from visibly labor-intensive industries like construction.
