Light weight Real Time Indian Sign Language Symbol Recognition with Captioning and Speech Output

Sign language is the principal mode of communication for the vision and hearing impaired. It has allowed us to communicate with differently-abled individuals in a manner that is as expressive as a spoken language. Sign language has satisfied all the objectives of a traditional language system, like expressing emotions while also fulfilling a greater humanitarian role. Just like there are different dialects of languages spoken around the world, there also exist dialects of sign language that differ from country to country. Each of these dialects use a different symbol or method of representing the same lexicon. This work proposes a method of dynamic Indian sign language translation. that uses neural networks and image processing. The parameter utilized is minimum with good degree of predicted accuracy is the major contribution of this paper. The trained weights for the model are small enough to be efficiently ported over to a mobile phone or web browser and can be used without an active internet connection. The model is tested on the dataset with 99% accuracy.