Asamyuta Hasta Mudras Recognition Using MediaPipe Keypoint Extraction and Random Forest Classification

Bharatanatyam is a dance of Tamil Nadu in southern India. This dance form includes the hand gestures known as the Mudras which holds immense symbolic meaning and play a vital role in storytelling. Each mudra carries its own essence, depicting various emotions, deities, and natural elements. We are here focusing on the Asamyuta hasta mudras or single hand gestures are expressions to be shown by a hand. This work addresses mudra recognition by framing it as a 28-class classification problem. This is achieved by utilizing MediaPipe for hand key points in a CSV format and demonstrates effective performance even with a limited dataset. This will achieve robust mudras recognition through a supervised machine learning model.