Modeling of Upper Limb and Prediction of Various Yoga Postures using Artificial Neural Networks

Regularly monitoring the movement and activity of a person is called activity recognition. Self-guided framework to practice yogasana makes individuals to learn correctly and practice yoga postures can be constructed using human posture recognition without taking help from others. We propose a method for effectively identifying and recognizing various yoga poses that makes use of “Artificial Neural Network” ANN algorithms. The selected dataset consists of 85 videos with 15 participants performing six yoga poses. The Media pipe library was used to extract the users’ key points. As a Artificial Neural Network (ANN) model, real-time-monitored videos have been used to identify yoga poses using an Artificial Neural Network (ANN) and Long-Short Term Memory (LSTM). In particular, the key point features are extracted by the ANN layer & subsequent LSTM layer recognizes sequence of frames in order to make predictions. The postures are categorized as wrong or right in the following: The system provides feedback in the form of text or speech if a correct pose is found. Data sets and machine learning are combined due to the synergy between the two fields as that machine learning methods, particularly ANN, can effectively recognize and make datasets interoperable.