<div dir="auto"><div dir="auto"><b>Machine learning to make sign language more accessible</b></div><div dir="auto"><<a href="https://blog.google/outreach-initiatives/accessibility/ml-making-sign-language-more-accessible/amp/">https://blog.google/outreach-initiatives/accessibility/ml-making-sign-language-more-accessible/amp/</a>></div><div dir="auto"><br></div><div dir="auto"><div dir="auto">
<div>
<div>
<div>
<div><i>Kemal El Moujahid</i></div>
<div><i>
Director, Product Management
</i></div>
</div><i>
Published Dec 01, 2021</i> </div>
</div></div></div><div dir="auto"><br></div><div dir="auto">Google has spent over twenty years helping to make information accessible and useful in more than 150 languages. And our work is definitely not done, because the internet changes so quickly. About 15% of searches we see are entirely new every day. And when it comes to other types of information beyond words, in many ways, technology hasn’t even begun to scratch the surface of what’s possible. Take one example: sign language.</div><div dir="auto"><br></div><div dir="auto">The task is daunting. There are as many sign languages as there are spoken languages around the world. That’s why, when we began exploring how we could better support sign language, we started small by researching and experimenting with what machine learning models could recognize. We spoke with members of the Deaf community, as well as linguistic experts, working closely with our partners at The Nippon Foundation, The Chinese University of Hong Kong and Kwansei Gakuin University. We began combining several ML models to recognize sign language as a sum of its parts — going beyond just hands to include body gestures and facial expressions.</div><div dir="auto"><br></div><div dir="auto"><b>...</b></div><div dir="auto"><b><br></b></div><div dir="auto">Advances in AI and ML now allow us to reliably detect hands, body poses
and facial expressions using any camera inside a laptop or mobile phone.
SignTown uses the <a href="https://google.github.io/mediapipe/solutions/holistic">MediaPipe Holistic model</a>
to identify keypoints from raw video frames, which we then feed into a
classifier model to determine which sign is the closest match. This all
runs inside of the user's browser, powered by <a href="https://www.tensorflow.org/js">Tensorflow.js</a>.<b><br></b></div><div dir="auto"><br></div><div dir="auto"><br></div><div dir="auto"><i>Click link for full article</i></div><div dir="auto"><i><br></i></div><div dir="auto">Mark Mandel</div></div>