Indian Classical Dance Recognition using Convolutional Neural Networks
Main Article Content
Abstract
The extraction and recognition of human movements in performing arts, particularly dance, represent a complex and captivating area of research. In the modern era of globalization, the creative expression and production strategies of classical dance have evolved significantly, creating a need for advanced technological systems to preserve and analyze these art forms.
The primary objective of this study is to develop an automated machine learning framework capable of detecting and recognizing the movements of Indian classical dancers from video sequences. Understanding the semantics of dance movements not only aids in safeguarding India’s rich cultural heritage but also supports the development of digital tools for dance education and performance analysis.
The proposed system begins with dancer detection in video frames using background subtraction and Histogram of Oriented Gradients (HOG) techniques. Subsequently, body part segmentation—including the hands, legs, and face—is carried out through HAAR-based and skin color detection methods. From each segmented frame, essential features such as Image Color Coherence Vector (ICCV), Gray Level Run Length Matrix (GLRLM), and Shift Invariant descriptors are extracted to interpret the dancer’s movements effectively. Finally, the system employs a Convolutional Neural Network (CNN) for the classification and recognition of dance sequences.