Here, we demonstrate the use of AI to unlock a 'voice' for the speech impaired, creating a novel way for people to speak silently and imperceptibly when in public or otherwise, opening a new medium of communication for people with speech disorders. By simply thinking of a number in your mind, our system can detect with high accuracy what your were planning to say within a low latency. Minute muscular activation signals occur when a person intends to speak, which we capture in real-time and map to the numbers. We argue our machine learning pipeline is robust and repeatable, and it reaches a classification accuracy better than 99% for the numbers 1-5 and provide a reproducible framework for further research.

• This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.

Creator

• Savoie, Luke
• Van Steyn, Jacob

Subject

• Well-Being
• Computing
• Inclusion
• Innovation
• Health

Publisher

• Worcester Polytechnic Institute

Identifier

• 106436
• E-project-042723-150345

Keyword

• Machine Learning
• EMG
• Human-Computer Interaction
• Subvocalization
• Biosensing

Advisor

• Yousefi, Ali

Year

• 2023

UN Sustainable Development Goals

• 3 - Good Health and Well-being
• 10 - Reducing Inequality

Date created

• 2023-04-27

Resource type

• Major Qualifying Project

Major

• Data Science
• Computer Science

Source

• E-project-042723-150345

Rights statement

• In Copyright

License

• All rights reserved

Last modified

• 2023-06-23