Bridging The Communication Gap: A Machine Learning Approach to Sign Language Recognition

A. Amulya, Afifa Butool, Sedmaki Aishwarya

Abstract: Sign language is the main and most effective communication tool that is through touch, among the hearing-impaired and the deaf. However, those that do not know it (through sign language) at all, therefore, a communication barrier exists in this situation. This project aims to build a strong system for sign language recognition based on Media Pipe, OpenCV, and Scikit-learn. Hand tracking in Media Pipe is used for visual hand detection and extraction of the pupil, which represents hand motion accurately and discriminately. OpenCV serves to offer comprehensive image processing such as scaling, normalization, and image data augmentation. This is done only after the preprocessing operation has ensured data best practices for training. The features that were acquired are then used to train the implemented machine learning models in Scikit-learn, for instance, Random Forests or Support Vector Machines for the classification of gestures.

Keywords: Hearing-impaired, Media Pipe, OpenCV, Random Forest, Gesture, Normalization