Hand Mudras and Indian Classical Dances: Techniques, Dataset and Future Direction

Concept, purpose and feelings of a dance is expressed through performer’s gesture. Proper interpretation of these signals requires skills acquired by different methods, artificial intelligence or human experts. In recent times, Dance Mudra Recognition is a very prominent and a multidisciplinary research topic. This paper discusses extensive research work carried out on various Indian classical dances by different researchers and automated recognition system developed by them. Author’s basic focus here is to study and discover an efficient intelligent method which will be capable of recognizing different hand mudras of Sattriya dance. A few efficient machine and deep learning methods are selected for working on our sattriya dance mudra dataset. The authors of this paper have used customized CNN model, ResNet50, ViT and YOLOV8 for training the data and were able to achieve very good accuracy. Till now no work is done using the above mentioned models for recognition of sattriya hand mudras.