EchoSign: A Mobile Application for Sign Language Recognition and Translation

Sign language recognition is an increasingly important yet underexplored area, particularly in the development of real-time communication tools for deaf and hearing individuals. This paper introduces EchoSign, a mobile application developed with Flutter that translates sign language gestures into both voice and text using a smartphone camera. The app features a robust machine learning model based on InceptionV3, which has been trained on a custom video dataset of 275 samples for each of 10 frequently used greeting words, as well as a supplementary Kaggle dataset for recognizing individual alphabet gestures. By leveraging these datasets, EchoSign effectively handles both word-based and alphabet-based sign language recognition, achieving an accuracy of 99% on the custom dataset. The real-time gesture recognition system processes video frames, predicts the corresponding gesture, and converts it to text while also offering text-to-speech functionality to vocalize the recognized words. This comprehensive approach ensures a fluid and accessible communication experience between signers and non- signers. By outlining the model architecture, dataset preparation, and performance evaluation, this paper highlights how EchoSign delivers a portable, real-time solution that can be easily deployed on modern mobile devices, offering a practical tool for breaking communication barriers and fostering inclusivity.