Inspiration
Back in high school, we had a friend who couldn’t speak but was incredibly bright. Despite his intelligence, he wasn’t allowed to study with us simply because he lacked a voice, not a brain. This experience inspired us to create a solution that ensures no one is excluded due to their inability to speak. Today, I challenge everyone to imagine having something to say but no way to express it or calling a friend on a video call and not being able to communicate. These are the barriers we aim to break. AI_SigniSpeak is more than a tool it’s a movement toward a more inclusive and equitable world. By addressing the needs of individuals with speaking disabilities, we are shaping communities where everyone has the opportunity to thrive, regardless of their abilities. This solution is not just a technological innovation it’s a lifeline for millions of people who have been excluded from virtual communication due to a lack of accessibility.
What it does
AI_SigniSpeak is a browser extension that enables real-time sign language to speech conversion, allowing individuals with speaking disabilities to express themselves on virtual platforms. It integrates seamlessly with tools like Zoom and Microsoft Teams, fostering inclusivity and breaking communication barriers.
How we built it
We built AI_SigniSpeak using cutting-edge AI technologies, including computer vision for sign language recognition and natural language processing (NLP) for speech synthesis. The extension is designed to be lightweight, user-friendly, and compatible with major browsers. We collaborated with deaf communities to ensure accuracy and cultural relevance.
Programming Language: We used "Python" for its robust libraries and frameworks, such as TensorFlow, OpenCV, and PyTorch, to develop the AI models.
- Sign Language Recognition: Leveraged computer vision techniques to detect and interpret sign language gestures in real time.
- Speech Synthesis: Integrated NLP and text-to-speech (TTS) technologies to convert translated text into natural-sounding speech.
- Browser Extension: Developed the extension using web technologies (HTML, CSS, JavaScript) to ensure seamless integration with platforms like Zoom and Microsoft Teams.
Challenges we ran into
Accuracy: Achieving high accuracy in real-time sign language recognition was a significant technical challenge, requiring extensive testing and optimization.
- Integration: Ensuring seamless integration with multiple virtual meeting platforms required extensive testing and collaboration with platform APIs.
- Localization: Adapting the solution for different sign languages (e.g., RSL, KSL) was complex but rewarding, as it involved understanding unique linguistic and cultural nuances.
- Affording a Programming Support Team: Building and maintaining a skilled team of developers and AI experts was a financial and logistical challenge.
- Access to Datasets: Acquiring high-quality, diverse datasets for training sign language recognition models was difficult due to limited publicly available resources.
- Training Models: Training AI models for real-time processing required significant computational resources and time, which posed a challenge for our team.
- Full-Time Support and mentorship: Providing ongoing support and updates for the solution while balancing other commitments was a challenge for the entire team.
## Accomplishments that we're proud of - Successfully developing a working prototype for real-time sign language to speech conversion.
- Partnering with local deaf communities to ensure the solution meets their needs.
- Creating a tool that has the potential to impact millions of lives across East Africa.
- Building a sample program that detects hand movements and accurately names the sign, demonstrating the feasibility of our approach. ## What we learned
- The importance of collaboration with end-users to build a truly inclusive solution.
- The technical complexities of real-time AI processing and how to optimize for low latency.
- How to adapt technology to diverse cultural and linguistic contexts.
- The value of putting ourselves in others' shoes to deeply understand their challenges and needs, ensuring our solution is both practical and impactful.
## What's next for AI_SigniSpeak
-Expanding support for more sign languages, including Ugandan and Tanzanian Sign Languages.
- Enhancing accuracy and reducing latency for real-time communication.
- Partnering with institutions and organizations to scale the solution across East Africa and beyond.
- Creating a digital platform designed by and for the deaf community, enabling live streaming, public speaking, and presentations.
- Developing a keyboard for blind people to help them write, learn, and use computers effectively.
- Building a dedicated supporting team to refine and perfect the solution, ensuring it meets the needs of all users.
Built With
- amazon-web-services
- browser-extension-apis
- docker
- firebase
- flask
- git/github
- google-cloud
- gtts)
- html/css
- javascript-(react.js
- jupyter
- keras
- mediapipe
- mongodb
- mysql
- notebooks.
- opencv
- postgresql
- python
- pytorch
- scikit-learn
- tensorflow
- webrtc)
Log in or sign up for Devpost to join the conversation.