Indian classical dance action identification using adaptive graph matching from unconstrained videos

Abstract

Extracting and recognizing complex human movements from unconstraint online video sequence is a challenging task. In this work the problem becomes complicated by the use of unconstraint video sequences belonging to Indian classical dance forms. A new segmenta-tion model is developed using discrete wavelet transform and local binary pattern features for segmentation. We also explore multiple feature fusion models with early fusion and late fusion techniques for improving the classification process. The extracted features were represented as a graph and a novel adaptive graph matching algorithm is proposed. We test the algorithms on online dance videos and on an Indian classical dance dataset prepared in our lab. The algorithms were tested for accuracy and correctness in identifying the dance postures.