Performance Analysis Of Supervised Classification Models On Gesture Recognition Of Single Handed Alphabet System Of Indian Sign Language
This paper studies the performance analysis of different supervised learning classification models in the context
of static gesture recognition of the single handed alphabet system, a subset of the Indian Sign Language, with leap motion
controller as the apex in recognizing hand gestures. Innovations in the field of machine learning and pattern recognition can
be seen prominently in speech recognition, which today can be achieved with high accuracy. But, for the hearing and speech
impaired individuals, who are a substantial part of the Indian population (2.78%), an efficient model to recognize the
patterns of the articulated hand gestures and convert them to speech, is yet to be found. This paper would especially focus on
improving the prediction accuracies of the different supervised classification models in recognition of these hand gestures
and would aim to extend the same for dynamic gesture recognition system. Emphasize has been given to take into
consideration a diverse dataset that would lead us to stable gesture recognition system with best possible accuracies, the
classification models being used are Support Vector Machines, K- Nearest Neighbors and Decision Tree. Finally, we would
evaluate the results of the different classifiers, analyze the features and the techniques that were devised to give such results.
Keywords - Indian Sign Language (ISL), Support Vector Machines (SVM), K-Nearest Neighbors (KNN), Decision tree (DTree),
leap Motion controller, machine learning (ML), Stochastic Gradient Descent(SGD), gesture recognition.