An Effective Machine Learning-based Segmentation and Feature Extraction Technique for Muscular-Disorder
Somashekhar S. Dhanyal, Suvarna Nandyal · 2023
Application of technology for the classification of sports and exercise, Yogasan identification is now common practice for numerous people in modern society especially when dealing with muscular disorders. Over the past three decades, research has focused on motivating and difficult problems related to the usage of computer vision for target recognition and identification. The skeleton parts of the head, neck, arms, hands, legs, and feet are designed and stored as a vector in this research to identify the human yoga asana pose. These skeleton elements are also responsible for head movements, hand postures, and standing postures. Every single interaction between the skeleton joints makes a pose, and every feature is described as a set of moveable joints. The weighted skeletonization-based image segmentation is proposed for the enhanced Yogasan recognition from the selected keyframes, the most relevant keyframes are selected using the Euclidean distance measure. The performance of the proposed method is analyzed by considering the features as well as the training percentage. The accuracy, precision, and sensitivity based on the features of the multi-SVM classifier are 89.020 %, 89.720 %, and 90.855 %, and the training percentage using the multi-SVM classifier is 94.268 %, 93.297 %, and 97.375 %.