Hi everyone Ignitious here!!! As an engineering student who’s always been fascinated by neural interfaces, I’m thrilled to be leading an exciting student project team to decode silent speech. My teammates Mohamed and Lakshmi and I have set out to develop a non-invasive system that can understand words solely from the subtle facial gestures that accompany imagined speech.
The key innovation in our approach lies in leveraging the OpenBCI Cyton board provided through their Discovery program. This versatile biosensing platform allows us to position medical-grade amplifiers over specific facial muscles involved in articulation. As the user thinks through pronouncing words internally, without any vocalization, the amplifiers can pick up on the resulting millivolt muscle activation signals. These can then be processed through optimized machine learning algorithms to classify speech patterns.
Our system has the potential to be truly life-changing for those robbed of voice from conditions like laryngectomies or paralysis. By decoding their intended speech directly from neuromuscular signals, we can give them a way to converse swiftly again. The goal is to correctly classify over 100 common words from facial electromyogram data with over 90% accuracy.
We intend to openly share our collected datasets with the research community. By allowing others to build readily on our methods and validation, we hope to catalyze innovation in non-invasive speech prosthetics. Assistive technologies that leverage silent speech decoding could help millions communicate more independently. Access to OpenBCI’s versatile Cyton board has already proven invaluable for my team by enabling rapid experimentation on a budget. We’re extremely grateful for such hardware donations that actualize ideas beyond textbook concepts. By empowering enthusiastic young minds, companies like OpenBCI pave the way for groundbreaking innovations we can’t even envision yet!
If you have any questions, suggestions, tips, Feel free to contact me at [email protected].
Leave a Reply
You must be logged in to post a comment.