HearMe

Inspiration

Our inspiration comes from personal experiences—each of us either has family members or close friends who struggle with speech disabilities. We realized that while voice assistants like Siri and Alexa have become an integral part of daily life, they remain inaccessible to many people with speech disabilities. This inspired us to create HearMe, a tool to bridge this gap and make technology more inclusive.

What it does

HearMe uses supervised learning to train a machine learning algorithm on a dataset of speech patterns from individuals with disabilities. The algorithm learns to recognize and translate these unique speech patterns into more standard speech. This translation allows the system to bridge the communication gap, enabling voice recognition technology to better understand people with speech disabilities.

How we built it

We built HearMe using TensorFlow and Python. The system is trained on a database of diverse speech patterns from individuals with different disabilities. We developed models capable of recognizing these patterns and translating them into standard speech inputs. Our backend is designed to interface with public databases from TORGO and putting them into our project, making it compatible with popular platforms like Siri and Alexa.

Challenges we ran into

One of the biggest challenges was finding publicly available databases that included both audio recordings of individuals with speech disabilities and matching transcripts. Many datasets either lacked diversity or didn’t include the types of speech impediments we were focusing on. This significantly slowed our initial development, as we had to spend time curating and preprocessing limited data. Additionally, fine-tuning our model to account for a wide range of speech patterns while ensuring real-time accuracy was a technically complex task.

Accomplishments that we're proud of

We're proud of creating a system that could change lives by making everyday technology accessible to those who have been left out. We’ve successfully trained a model that can recognize complex speech patterns with impressive accuracy. Additionally, we’ve maintained a focus on keeping the project open and accessible, aiming to make this technology available to anyone who needs it.

What we learned

We gained valuable experience using TensorFlow, and Python. Working as a group of three in a 36-hour hackathon taught us the importance of teamwork, communication, and quick decision-making under pressure, along with delegating work to each other. Most importantly, we learned how crucial it is to prioritize inclusivity when developing technology, ensuring that our solution addresses the needs of underrepresented groups.

What's next for HearMe

We aspire to be pioneers in creating larger, more comprehensive databases of speech patterns from individuals with disabilities. By building and sharing these datasets, we hope to enable future teams and innovators to continue advancing accessibility for underrepresented groups. Our vision is to lead the way in making these technologies more inclusive, empowering not only voice assistant users but anyone relying on speech-to-text systems across the globe.

Built With

• figma
• jupyter
• python
• tensflow