The problem we are trying to solve
For millions of people around the world who have difficulty speaking, face-to-face communication can be very challenging.
Using voice activated technologies can be frustrating as well. While tools like Google Home or the Google Assistant can help people call someone or adjust lighting or music, they may not work as efficiently or accurately for those with impaired voices. This is due to the fact that speech recognition algorithms have seen millions of examples from people whose voices sound “typical,” but not many examples from people with speech impairments.
Ultimately, we hope to improve peoples’ independence and ability to communicate by making technology work better for everyone. Our first goal is to gather enough voice recordings from people with speech impairments to enable the development of better speech recognition models.
How our research works
The first step of our research effort is to record voice samples from contributors who have non-standard speech. We ask volunteers to read specific phrases, so we have transcriptions (or labels) matching the audio. Our team of research scientists then uses this data to refine our speech recognition algorithms to better understand different types of speech.
Research timelines can be long and unpredictable; sometimes years long. But if you would like to embark on this journey with us, we would love to have you.
Learn more about how on-device speech recognition at Google works and what techniques have been explored for impaired speech (Parrotron and Euphonia):
Who we are
Photo of many Project Euphonia team members as of June 2019
We’re a group of research scientists, software engineers, designers and product managers at Google dedicated to making technology work better for everyone.