STATIC SIGN LANGUAGE RECOGNITION USING IMAGE PROCESSING AND DEEP LEARNING TECHNIQUES

Abstract

The deaf and mute community has a great difficulty articulating their thoughts and opinions to everyone else; sign language is their most eloquent means of communication, but the general public is unaware of sign language, making it difficult for the mute and deaf to communicate with others. To address this communication gap, a system that can accurately convert sign language gestures to speech and likewise in real-time is required. This work proposes Effi-CNN, an image Sign Language Recognition (SLR) system. Our system uses transfer learning with EfficientNetB2 as the basic model to transform sign gesture photographs to words. We've also created a system that translates hand movements into text instantaneously. We evaluated our on eight publically available datasets, including the Massey University gesture dataset, ArSL2018 dataset, MNIST-ASL dataset, and others. Comparing our results to state-of-the art algorithms, the experimental findings show that our technique is more successful. The results show that our Effi-CNN surpasses most of current existing solutions, and it has the ability to categorise a large number of gestures with a low rate of error.