Page 141 - AI for Good Innovate for Impact
P. 141
AI for Good Innovate for Impact
(continued)
Item Details
Testbeds or Pilot Deploy- We will use the facilities at the Media Research Laboratory of our
ments university. Finally, we intend to test in a typical home environment. 4.1-Healthcare
Code repositories STGCN: [8], Mediapipe: [9]
2 Use Case Description
2�1 Description
Our solution is one that harnesses AI in assisting individuals suffering from musculoskeletal
disorders, particularly in regions where there exists a significant shortage of physiotherapists.
Therapeutic exercises are frequently prescribed as part of rehabilitation; however, in India,
there is an average of merely 0.36 physiotherapists per 10,000 individuals, highlighting an
urgent need for a digital solution. This use case presents a system designed to deliver real-
time corrective feedback during physical therapy exercises [5]. The proposed system employs
a simple camera and a Raspberry Pi as in Figure 1 utilising MediaPipe for pose detection and
XNNPACK to accelerate inference. The system compares live user movements against reference
joint angles and variations in distance between major key points [1]. The matching process is
further refined by incorporating user-specific factors such as age and historical performance to
ensure a personalised experience. Feedback is conveyed in textual, auditory, and visual formats
via a display connected to the Raspberry Pi. Additionally, a bespoke Spatio-Temporal Graph
Convolutional Network (STGCN) [2], hosted on the cloud, evaluates the performed exercises
and generates an assessment score as in Figure 2. This score is displayed to assist users in
monitoring their progress. Hence, it is safe to say our solution holds the potential to democratise
access to expert guidance and corrective feedback, making at-home physiotherapy both more
accessible and cost-effective.
105
