|
Title:
|
Cross-Cultural Modal Fusion Recognition of Dance using Multimodal Deep Learning Framework
|
|
Author:
|
Dr Dheerdhwaj Barak
|
|
Keywords:
|
Multimodal Deep Learning, Skeletal Motion Analysis, Cross-Cultural Dance Recognition and Cultural Heritage Informatics and Explainable AI (XAI).
|
|
Abstract:
|
Multimodal Deep Learning framework for Automated Recognition and classification of movement patterns for Cross-cultural dance fusion (CCDF) of Indian classical and western style dance. The model is capable of learning and fusion of three different modalities: skeletal motion trajectories, rhythmic audio features, and semantic gesture annotations. It is built around spatio-temporal analysis which allows the learning of disentangled and characteristic motion-signatures from each tradition. To make the architecture interpretable, attention mechanisms are incorporated which visualise the focus of the model to the diagnostically meaningful movements and rhythmic cues. The proposed method was applied to a carefully selected dataset, and the results showed that the method is able to detect the fusion of choreography at a high rate of performance (as measured by a novel fusion index). This work makes a significant contribution to the areas of dance informatics and digital heritage as it gives an analytical tool to understand cultural-artistic synthesis in performing arts, which can be scaled up.
|
|
Download Paper:
|
|
|