SIGNLANGUAGETOSPEECHCONVERTERFORINDIANLANGUAGES USING NEURAL NETWORK

Author (s) : Dr.V.Vijayakumar, Dr.S.Muniyappan, J.Shanbagam, M.Brindha

---

Abstract :
Sign language is an extremely importantcommunication tool formanydeaf and mute people. So we proposed a model to recognize sign gestures usingYOLOv5 (You only look once version 5). This model can detect sign gestures incomplex environment also. For this model we got the accuracy of 95.7% withprecision of 86.6% and recall of 90.2%. The proposed model has evaluated on alabeled dataset Roboflow. Additionally we added some images for training andtesting to get better accuracy. We compared our model with CNN (convolutionalneural network) where we got accuracy of 91.98%. We checked this model forrealtimedetectionalsoandgottheaccurateresults.An efficient technique has been developed to let silent people interact with regular people effectively. Our goal is to create a system that can examine and identify different alphabets from a collection of sign images. The application employs a number of image processing techniques, including feature extraction and segmentation, to achieve this. Convolutional Neural Network is a machine learning technology that we employ for detection. Verbal Communication is the only way using which people have interacted with each other over the years but the case stands different for the disabled. The barrier created between the impaired and the normal people is one of the setbacks of the society. For the impaired people (deaf & mute), sign language is the only way to communicate. In order to help the deaf and mute communicate efficiently with the normal people, an effective solution has been devised. Our aim is to design a system which analyses and recognizes various alphabets from a database of sign images. In order to accomplish this, the application uses various techniques of Image Processing such as segmentation & feature extraction.

Keywords : Indian Sign Language Recognition, Translation, Feature Extraction, Convolutional Neural Network, Neural Network.

Conference Date : 17/09/2021

Pages : 156

Paper ID : IESMDT154