Bharatanatyam dance classification with rough set tools

Abstract

BharataNatyam (BN) Choreography is known to be an intuitive domain. We have attempted to aid the choreographer for this creative domain of choreography for pure dance movements (used for aesthetics) called Nritta with our ArtToSMart (System Modeled art) system. Various automated techniques have been used for Western dance notations, its animation as well as for choreographic skills; but we have not been able to find enough on BN. We have used rough set tools to build a classifier for the dance poses, since manual methods are subjective and time-consuming. Thirty attributes have been used to define the human body. We obtained eight reducts from among these 30, and also decision rules which frame the grammar of a BN pose. The results are promising and have higher accuracy for about 500 instances. A comparative study has been done with two different tools namely WEKA (Waikato Environment for Knowledge Analysis) 3.6.11 and RSES (Rough Set Exploration System) 2.2.2 which has given 87.42% and about 76.8% classifier accuracy, respectively.