Bi-Directional Communication Tool for Indian Sign Language to Text/Speech

The sharing of information through communication is an essential approach in human communication, but patients with hearing loss encounter difficulties being understood by people without knowledge of sign language.This paper proposed a deep learning model for Indian Sign Language (ISL) Recognition and Text-to-Speech (TTS) for narrowing this gap in communication.The system aims at detecting ISL alphabets using hand gestures and converting it into the textual/messages with the sound maker to have an output that will support two way communication.The model uses convolution neural networks (CNN), which benefit from prior training structures such as ResNet-18 to achieve precise gesture recognition.ISL The dataset was trained and used for real-time indications for practical use.The project also presented a strong assessment framework to evaluate performance metrics, including accuracy, accuracy, recall and F1 score.Visualization techniques such as Confusion Matrix give insight into the efficiency of the model.The system shows great accuracy in data sets and has the option for practical applications, such as paying aid to deaf and hearing -affected communities, learning sign language and publication in public places such as hospitals or offices.Future improvements include expanding hand vocabulary, real -time sentence recognition, cross-language support and integration with portable devices.Human-written text: Through technology to increase approachability and decrease boundaries of communication, this project represents an important stage towards creating a positive society.