Indian Sign Language Recognition with Conversion to Bilingual Text and Audio

Communication is an important part of our day-to-day life as it allows us to express ourselves easily. However, speech isn’t the only way for successful communication, gestures and signs also play an equally meaningful role in communication. Indian Sign Language Recognition is a topic that is far less researched as compared to its counterparts such as American Sign Language Detection. In this paper, signs of the alphabet and digits were captured using an open cv and combined with dataset images obtained from Kaggle. The dataset was processed by resizing, flattening, and converting the images into binary. Though existing systems identify Sign Language with sufficient accuracy, this system incorporates the recognition from a very easy-to-use keyboard interface as an assistant/educator having a feature focused towards Indians that will allow getting the desired result in the text as well as audio format. For displaying the results, two languages were chosen: English & Hindi. Out of the four models trained, namely SVM, KNN, Random Forest, and Decision Tree classifiers; the Random Forest Classifier outperformed other models and it was the one to give highest accuracy of 99.6%.