International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 06 | Jun 2025
p-ISSN: 2395-0072
www.irjet.net
A Real-Time Physiotherapy Pose Monitoring and Feedback System Using Machine Learning Prof. Jaitee Bankar1, Jay Patil2, Atharv Chormale3, Chaitanya Pathak4. 1Assistant Professor, 2,3,4Student. IT Department, Savitribai Phule Pune University, Sinhgad College of Engineering, Warje. --------------------------------------------------------------------------***-----------------------------------------------------------------------
visual and audit feedback via a custom TKINTER-based GUI. This allows users to modify their attitudes while they are running. It enables patients with implementable feedback, promotes consistency in rehabilitation and reduces reliance on continuous therapist monitoring.
ABSTRACT Domestic physical therapy exercises often lack professional supervision, which can slow recovery or lead to further injuries. This project uses computer vision and machine learning to provide real-time pose recognition and feedback systems. This is specialized for general rehabilitation exercises. The system uses directional points like media pipe taylors in webcam inputs and uses logistics regression models to classify physical therapy poses such as wrist stretching, grip movement, shoulder roll, fingertip touch, cat child, vertebral container, and arm rotation. A graphical user interface (GUI) created with TKINTER displays recognized poses along with trust values and provides realtime correction feedback via visual information and audio requests. The system provides accurate and immediate attitude assessments and improves the security and effectiveness of unattended rehabilitation. This work shows what is scalable and accessible.
2.FUTURE SCOPE The proposed system can be extended to a mobile application to enable users to perform physical therapy exercises that conveniently perform smartphones using their smartphones. Future versions can integrate languagebased feedback and enable hands-free interaction. Additionally, the integration of portable devices improves pause recognition accuracy and allows for continuous monitoring. The system can also be developed to include report counts, personalized athletic planning, and performance tracking. These improvements allow therapists to monitor patient progression from afar and more effectively adapt rehabilitation protocols.
KEY WORDS:
3.LITERATURE REVIEW
Physiotherapy, pose detection, holistic media pipes, logistics regression, real-time feedback, human pose estimation, home rehabilitation, computer vision, TKINTER-GUI, posture correction, machine learning, rehabilitation exercises, landmark tracking.
Physical therapy pose detection recorded different advances in different modalities. Agrawal et al. (2020, Arxiv) Tensorflow-based skeletal pose estimates for custom data records (5,500 images) reach 99.04% accuracy in random forests, although limited by various data. Verma et al. (2020, arxiv) Proposal *Physiotherapy82 *, 82 Poznan large hierarchical data set using Denenenet-based CNNs, improving classification robustness under occlusion and dispersion. Anantamek (2019, IEEE) used EMG signals to detect posture adjustment with lower extremities. This provided an accuracy of 87.43% reporting and highlighting accuracy at the muscle level despite sensor complexity. Balakrishnan and Zhao (2020, DSPACE-MIT) introduced modularly generated neuronal networks for new pose integration on activities and focused on realism through controversial training. Gochoo and Tan (2018, IEEE) developed an IoT system using eight raw thermostunts and deep CNNs, and designed privacy with a F1 score of 0.9989 in LaTENCY (107 ms), ideal for domestic-based applications. Gregory (2020, IEEE) has developed a real-time system for detecting child posture accommodations that support coaches with standardized feedback, although limited by
1.INTRODUCTION Physiotherapy is a key component of the recreational and rehabilitation process after musculoskeletal injury, surgery, or chronic physical illness. A significant portion of today's physical therapy is performed at home, and patients are expected to perform regular and accurately prescribed exercise. Without professional monitoring, maintaining proper form can be difficult to increase the risk of ineffective treatments and other injuries. The system supports commonly defined physical therapy exercises, including wrist tracks, grip movements, shoulder rolls, fingertip touches, cat cows, spine intelligence, and arm rotation. These exercises aim for joint flexibility and joint flexibility and strength nuclear areas in orthopedic and neurological case rehabilitation protocols. These sights are used by light but effective logistic regression models to classify poses in real time. This system provides immediate
© 2025, IRJET
|
Impact Factor value: 8.315
|
ISO 9001:2008 Certified Journal
|
Page 1131