Sign Language To Categorical Text using Convolutional Neural Network

Abstract

Recognizing hand movement is very important in Sign Language detection. In Proposed paper Indian Sign Language (ISL) detection using Convolutional Neural Network (CNN) is used. RESNET 101 is used for representation of best features in image dataset. Indian Sign Languages has 0–9 digits and 26 alphabets. The working architecture model used in project contains input layer which is used for representing image in the form of pixels, activation layer for training the model to learn from different data, pooling layer for reducing the computational time by optimizing the dimensional space of each feature map, flatten layer to decrease the array dimension and dense layer is used in last layers where it gathers the data from multiple neurons. To prevent the overfitting dropout layer is used. Sign Language is used to express emotions, sentences, words, alphabets and numbers. The hearing and speech impaired people are struggling to express their emotions to normal people. The motivation behind developing such vision based system which can detect and recognize the hand movements and convert it into text, was to bring the hearing and speech impaired people to the normal community where they can express themselves easily using an economical and efficient application. In the paper gathering of data, design workflow and Convolutional Neural Network is discussed.