This video interview features MSR’s Jasha Droppo:
I have been with Microsoft Research since July, 2000. My primary task has been to explore different techniques to make ASR more robust to additive and channel noise. Other projects I’ve worked on include general speech signal enhancement, pitch tracking, multiple stream ASR, novel speech recognition features, MiPad multimodal interface, cepstral compression and transport, and the WITTY microphone.
I really couldn’t believe what I was seeing. Then, I couldn’t believe what I was hearing. If you can’t see the embedded YouTube clip, you can either download the MP4 or watch the video on the Web:
Argh! Release the code, already – don’t make me have to stage a podcaster protest, please? And as if this wasn’t enough to frustrate you audio jockeys, there’s another MSR team that has developed an application that will actually edit out speech-to-text recognized words on-the-fly. You could effectively get rid of all those “uhhs” and “ahhs” with simple clicks. I didn’t catch their names, however.
Please, would someone at Microsoft just… put this stuff out there and let us (your users, not committees) decide whether it’s worth pursuing?!
[tags]microsoft, ASR, MiPad, WITTY, noise cancel[/tags]