On the Classification of Kathakali Hand Gestures Using Support Vector Machines and Convolutional Neural Networks

• Main Authors: Bhavanam, Lakshmi Tulasi, Neelakanta Iyer, Ganesh
• Format: Conference Proceeding
• Language: English
• Subjects: Clustering algorithms; Convolution Neural Network; Deep Learning; Feature extraction; Histograms; Image Processing; Image segmentation; Kathakali; Machine learning; mudra; Support Vector Machine; Support vector machines; Videos
• ISSN: 2640-5768
• DOI: 10.1109/AISP48273.2020.9073398

Indian classical dance such as Kathakali is composed of complex hand gestures, body moments, facial expressions and background music. Due to the complexities involved in its hand-gesture language, it is often difficult to understand kathakali mudras. In this paper, we generated a dataset for Kathakali hand gestures and explore ways to recognize Kathakali dance mudras performed by artistes with the help of machine learning and deep learning techniques. There are 24 classes of hand gestures that are used to convey the story by the performer in Kathakali. We proposed a Support Vector Machine (SVM) model and Convolutional Neural Network (CNN) model which classify the images into 24 different classes. We compared the performance of machine learning algorithms and deep learning algorithms. Our results show that deep learning algorithms gave up to 74% accuracy. To the best of our knowledge, this is the first attempt to generate a dataset of Kathakali hand gestures, explore data pre-processing techniques for machine learning techniques and applying deep learning techniques for classification of Kathakali hand gestures.