Recognition of Yoga poses through an interactive system with Kinect based on confidence value

Abstract

Nowadays, the recognition of poses is a field of investigation that takes incredible significance for oneself preparing in different sports. Kinect offers a low-cost solution for the recognition of Yoga poses due to body tracking and depth sensor. In this research, we propose an interactive system for perceiving a few postures for learning Yoga that will be characterized by a level of trouble and coordinated with command voices to envision the guidelines and pictures about the stances to be execution. Likewise, posture correction instructions will be displayed for the user in real time made by an expert yoga trainer. Besides, the recognition algorithm is based on Adaboost algorithm in order to get a robust database for detecting 6 Asana Yoga poses. All data were obtained and analyzed according to the confidence which showed a maximum average value of 92%.