Standing Posture Rectification System

Lim, Jun Chian (2020) Standing Posture Rectification System. Final Year Project (Bachelor), Tunku Abdul Rahman University College.

Text Lim Jun Chian.pdf Restricted to Registered users only Download (3MB)

Abstract

It has become increasingly common for modern people having poor standing postures. With a long time of poor standing posture, it causes some symptoms such as difficulty breathing, chest tightness, indigestion, palpitations, etc. Therefore, a standing posture recognition system has been developed to rectify the improper standing postures of the user by alert him/her. In this project, the specifications and advantages of data acquisition are discussed. A correct and incorrect standing posture might look similar to each other with human eyes. Therefore, a posture recognition algorithm dedicated to recognizing and differentiating between a proper and improper standing posture has been developed. Research and study have been done at work related to the posture recognition system done by other authors. Besides, the methods and solutions of research used are compared and tabulated. In this project, Kinect was proposed as data acquisition for utilizing the skeleton information. However, the software used is MATLAB, and it used to train and implement the posture recognition algorithm. A comparison and research of machine learning algorithm are performed between SVM and ANN algorithm. With the result, it shows that the SVM with RBF kernel has the highest accuracy compared with others, where the accuracy is 97.9335%. Although the accuracy of SVM with the linear kernel is approximately same with RBF kernel, the input space of training and testing sets are not linearly separable. Thus, the linear kernel is speculated that the recognition system is biased. Besides, although the accuracy of the artificial neural network exceeds expectations, its accuracy is relatively moderate compared to other classifiers. Upon detecting an improper standing posture, the system will display ‘bad posture’ in computer to alert the user to rectify his/her standing posture. The purpose of this project is to focus on the posture recognition algorithm, the system is successful and accurate posture recognition. Since the purpose of this project is to demonstrate the idea, other parts will not be studied, especially the hardware. Three objectives were achieved in this project.