Tracking the DNA of Sound with MATLAB
Ken Sutton, Yobe
Startup Yobe has created software that can accurately track a voice’s “DNA” in any auditory environment and separate a voice of interest from background noise, including other voices. Their approach uses MATLAB®, signal processing, and artificial intelligence to improve the overall sound quality and speech command accuracy for far-field applications and speaker identification platforms. An example use case is the ability to use a low-power “wake word” (i.e., Alexa, Siri, OK Google) to validate authorized users, retrieve their user profiles, and preload their device and application settings in noisy environments.
As part of the MathWorks Startup Program, Yobe has access to MATLAB at a startup-friendly price and engineering support from MathWorks experts. MATLAB enabled them to demo their concept at every stage of development, helping them share their progress and build excitement for their software long before it was commercial grade.
Published: 3 Sep 2019
Imagine a world where you can speak to your device from across the room, and there's background noise. What Yobe allows us to do is create an environment where the signal processing can track your voice in any auditory environment-- basically signal processing that thinks.
When we started with the Yobe, what we had was the science. We had to put together demonstrations rather quickly, and that was done in MATLAB.
Yeah. So far Yobe, one of the advantages of using MATLAB is it's the language that a lot of our engineers are familiar with. So there's a language and signal processing, and that language is built on MATLAB. For us we'll always develop in MATLAB at this point.