Sign Language Recognition u sing Convolutional Neural Network and Long-Short Term Memory

The system ”Sign Language Recognition using Convolutional Neural Network(CNN) and Long-Short Term Memory(LSTM)”aims at building a Deep Learning model using Computer Vision to recognize various signs used in American Sign Language(ASL). This system helps to establish a communication edge and understanding of a common language. People with some hearing disabilities need a special sign language to communicate. The system intends to reduce these difficulties. Our project consists of an automatic recognition system for hand gestures in ASL and furthers its conversion to text and speech. The idea deals with images of bare hands, which allows the user to interact with the system in a natural way. It provides an opportunity for visually im- paired people to communicate with normal people without the need for an interpreter. In this autonomous model classification, Deep Learning algorithms are trained using a collection of image data and tested using validation image dataset. This image dataset of different signs in ASL is downloaded from Kaggle and loaded to be used by the model. For the image dataset, binary images are used, which gives better results than the datasets used in previous models. Two Deep Learning algorithms are applied to the dataset which includes a CNN for classification and a LSTM network to store temporal features of the data used for sentence formation. The model specified in this article is used for recognizing gestures of A-Z alphabets, 0-9 digits, and also forms sentences using the recognized gestures of alphabets