Mathematics and Computer Science, Volume 1

21Pose Estimation Using Machine Learning and Feature Extraction

J. Palanimeera* and K. Ponmozhi

Department of Computer Apllications, Kalasalingam Academy of Research and Education, Krishnankoil, Tamil Nadu, India

Abstract

The recognition rate, robustness, and operational efficiency of machine vision-based yoga action detection of human postures is all typically low. The subtlety of the heritage and the diverse yoga asanas, blockages, and self-closures make contributions to this. This paper proposes a function extraction approach that mixes directional gradient of depth function (D-GoD) and local distinction of depth function (L-DoD) and employs a unique method that consists of eight close by factors round a picture element for joint evaluation for you to calculate the distinction among the picture elements for you to resolve this problem. This paper presents a method for reliably recognizing diverse Yoga asana using machine learning methods. A dataset comprising seven yoga asanas (Pranamasana, Dhanurasana, Dandasana, Gomukhasana, Garudasana, Padmavrikshasana, and Padmasan) was constructed using ten people (five men and five women) and a standard RGB webcam. The random-forest classifier is then professional within the use of the real-time statistics set. The random-forest classifier then educated the use of the real-time information set, and the picture elements on diverse areas of the human frame intensity image categorized the use of a random forest two-way vote casting procedure. ...