Sign Language Alphabet Recognition Using Convolution Neural Network
Mayand Kumar, Piyush Gupta, Rahul Jha, A. K. Bhatia, Khushi Jha, Bickey Kumar Shah
Abstract
Sign Language plays an indispensable role in the lives of people having speaking and hearing disabilities. Recognition of American Sign Language using Computer Vision is very challenging due to its increasing complexity and high intraclass variations. In this paper, convolutional neural networks (CNNs) are used to recognize the ASL Alphabets. This algorithm is useful to recognize it as a deep network, which is expected for the ASL alphabet classification task. Pre-Processing steps of the MNIST dataset are done in the first phase. After the first phase, different important features of pre-processed hand gesture image are computed. In the final phase, depending on the properties computed or calculated in the initial phases, the accuracy and AUC score of the network model with which it can recognize the sign language Alphabets were detected. The proposed CNN network has achieved an AUC score of 0.9981 and an accuracy of 0.9963.