Real Time Feature Based Finger Sign Language Recognition Using Deep Learning

Abstract

Sign language recognition is one of the fastest growing research areas today. Most of the research on HCI gesture recognition is based on artificial neural networks (ANN) or hidden Markov models (HMM). There are many effective algorithms for segmentation, classification, pattern matching, and recognition. The main goal of this paper is to compare the classifiers, which will definitely help researchers to find the best solution. The most important thing in gesture recognition system is the selection of input function and classifier. To improve the recognition rate and make the recognition system robust to viewpoint changes, the concept of shape descriptors from the available feature set is introduced. K-Nearest Neighbor (KNN), Proximity Support Vector Machine (PSVM), and Naive Bayes are used as classifiers to recognize static words. The performance analysis of the proposed methods is presented along with experimental results. Comparative analysis of these methods with other popular techniques demonstrates good real-time efficiency and robustness. The experimental results demonstrate the effectiveness of the proposed work, with the recognition efficiency of the KNN classifier being 78%, the PSVM classifier being 91%, the Naive Bayes classifier being 93%, and the proposed deep learning classifier being 97%.