查看文章

Recognition of sign language by hand gestures is one of the classical problems in computer vision. Conventional tools used for sign language translation involves application of linear classifiers such as kNN and Support Vector Machine (SVM)to perform the classification of hand gestures images. However, these methods require sophisticated features for classification. To automate the feature extraction and feature selection procedure, a Deep Neural Network (DNN)based machine translation is proposed in this work. Here, the images of English Sign Language (ESL)are identified using Deep Learning (DL)approach. A custom DNN with three convolution layers and three Max-Pooling layers are designed for this purpose. A top validation accuracy of 82% was obtained for the DNN structure proposed in this paper.