Interpreting Live Video Sign Language and Translate into Text

American Sign Language (ASL) is a method of communication developed for individuals who are deaf. While there are over 4 million individuals in the deaf community, around 500,000 individuals understand ASL. This research will explore machine learning techniques and a camera to interpret live video sign language and translate the ASL lexicon into text. This is performed through MediaPipe object detection and Open Computer Vision (Open CV) to track a user's hand and body movements and convert the footage into numerical values. These values are then compared to the dataset of the ASL alphabet that we have prepared, through 4 machine learning techniques. These techniques include Sequential Model, Random Forest, Naive Bayes and K-Nearest Neighbors (KNN) algorithms and are used to create machine learning models. The results suggest that KNN is the most accurate model (reporting accuracy up to 94%), where the optimal K value was 1. Random Forest tests accurately (up to 87%) when implemented and tested.