Rehabilitation Posture Correction Using Neural Network

Rehabilitation treatment is the process which helps people to restore their functioning. Nowadays poor body posture and forward neck problem has become a greater issue all around the world. At home or in the workplace, everyday activities such as force and repetition can play a part in causing these problems. Many people have this postural abnormality, which is most common in aging, working, and old-aged groups. For this, a low-price, reliable, and precise system for in-home rehabilitation is required. In this paper, the device for an in-home rehabilitation is designed for patients especially outpatients (who can’t visit daily or afford a personal trainer) for correcting their improper postures by giving visual feedback of postural data in real-time using LabVIEW software. Our system analyzes the patient’s whole-body posture by extracting the human skeleton called skeleton tracking which gives positional information for the 20 prime skeleton’s joint points that make up the human body using a Kinect sensor. We designed the Pattern Recognition Neural Network algorithm using MATLAB script function in LabVIEW to examine the tracked human skeleton. To verify that the posture is accurate or not, we have collected the data about 896 samples of exercising postures for both conditions to train the network. We performed 70% training, 15% validation and 15% testing of the network which results in classifying the right class of postures. Further testing will also be done.