Özet:

Abstract There has been a lot of research put into the statistical study of human behavior and movement. The ability to infer behavior from a single picture or a series of photos is a hot research area right now. Human Posture Recognition is a significant breakthrough in the direction of behavior comprehension since it may be used to identify actions taking place in a picture. Human posture estimate from the video is crucial for a wide range of uses, including the measurement of workouts, the identification of signs, and the manipulation of whole bodies via gestures. It may serve as the foundation for many dance, fitness, and yoga practices. It may also make augmented reality possible, where digital data is superimposed on the actual environment.The purpose of this study is to investigate and evaluate the viability of using Machine Learning to categorize human body position alongside a wide variety of complicated physical activities. Different basic, boosting, and ensemble machine learning methods are used in this study to categorize human posture based on the positions of individual body components (Distribution-aware coordinate representation). This study's dataset has 10 distinct physical positions that may be used to categorize 5 distinct workouts. These routines include variations on the push-up, pull-up, sit-up, jumping Jack, and squat. The final states of each exercise (the "up" and "down" postures for push-ups, for example) have been represented by two distinct classes. The strong predictions offered by the ensemble techniques were the result of the aggregation of the efforts of many different learners, making them more flexible.

Anahtar Kelimeler: