Sign Language Digit Recognition Using Different Convolutional Neural Network Model

Abstract

An enormous number of world populations in current time are unique in that sense that they have no broad language because of the absence of their hearing capability. The people with hearing impairment have their own language called Sign Language however it is hard for understanding to general individuals [1]. Sign digits are additionally a significant piece of gesture based communication. So a machine interpreter is important to permit them to speak with general individuals. For making their language justifiable to general individual’s computer vision based arrangements are notable these days. In this exploration of work we target to develop a model based on CNN to deal with the recognition of Sign Language digits. A dataset of 10 classes is used to train (70%), validation (20%) and test (10%) of the network. We consider three different models of CNN network to train and test the accuracy of sign digit. Among the three model transfer learning based pre-trained CNN performs better with test accuracy of 92%.

Sign Language Recognition Method Based on Palm Definition Model and Multiple Classification

“…The scope of machine learning applications is very diverse. A number of scientific directions that are used in gesture recognition tasks can be distinguished: classical learning [5][6][7][8][9], ensembles [10][11][12], neural networks, and deep learning [13][14][15][16][17][18][19][20][21][22][23][24][25][26][27].…”

“…Deep neural networks are the first to learn how to recognize gestures, one of the most complex objects for AI. They do this by breaking them into blocks, identifying the dominant lines in each, and comparing them to other images of the desired object [12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27].…”

Ensemble Transfer Learning for Hand-sign Digit Image Classification

“…As hand signs continue to play an increasingly prominent role in our daily lives, researchers have developed various techniques to enhance their identification. Among these, Convolutional Neural Networks (CNNs) have emerged as one of the most popular and effective approaches for extracting information from images of hands [4][5][6][7][8].…”