VocalZoom combines speech recognition with optics, reading vibrations on the face. That duo can determine when someone is speaking (and when they're not), making it easier to isolate speech out away from background noise.
The company got good press at CES this year (announced 3M, Motorola, and OurCrowd as investors) and was written up on the Economist:
http://www.economist.com/news/sc...
Gift cards for stocks? Stockpile hopes to get people excited about investing again