Skeleton Tracking Solutions for a Low-Cost Stroke Rehabilitation Support System
Summary
This study compared motion tracking systems for stroke rehabilitation. MediaPipe demonstrated high alignment with Kinect, offering a promising, low-cost alternative for assessing upper extremity movement quality.
Area of Science:
- Rehabilitation Engineering
- Biomedical Signal Processing
- Movement Science
Background:
- Computer-based motion assessment systems aid stroke survivors' home rehabilitation.
- Developing engaging, low-cost, and minimally complex systems is crucial for widespread adoption.
- Accurate assessment of upper extremity (UE) movement quality is vital for effective stroke recovery.
Purpose of the Study:
- To compare the accuracy of OpenPose and MediaPipe skeleton tracking for UE quality of movement assessment in stroke survivors against the Microsoft Kinect.
- To evaluate the feasibility of using OpenPose and MediaPipe as alternatives to Kinect for telerehabilitation.
Main Methods:
- A dataset of 15 stroke survivors performing rehabilitation exercises was collected.
- Three skeleton tracking systems (Microsoft Kinect, OpenPose, MediaPipe) were utilized.
- Root Mean Squared Error (RMSE) was computed to assess trajectory and kinematic variable alignment between systems.
Main Results:
- MediaPipe World Landmarks showed high alignment with Kinect-based assessments.
- OpenPose and MediaPipe data, when used with classification models, demonstrated accurate exercise performance assessment compared to Kinect.
- Kinematic variable alignment indicated MediaPipe as a viable alternative to Kinect.
Conclusions:
- MediaPipe presents a potential low-cost and effective alternative to Microsoft Kinect for upper extremity movement assessment in stroke rehabilitation.
- Minimal technical setup systems like MediaPipe can support autonomous rehabilitation exercises for stroke survivors.
- Further research should validate MediaPipe's efficacy across diverse stroke populations and exercise protocols.
