Saturday, March 15, 2008

Windows Vista speech recognition failure - the whole story



Here is what microsoft team is saying about this:
This specific problem looks like it was likely caused by a serious audio gain bug in the specific build that Shanen was using for the FAM (Financial Analysts Meeting) demo. I'm actually going to be taking a look at the machine personally this morning to see if that's what it was or if it was something else. No matter what it was, we'll address it before we ship in one way or another.

The results Shanen got during the demo were not at all similar to the results he was getting earlier yesterday morning, so there's certainly something else going on. If it always performed this poorly, we wouldn't ship it! :-)

I consistently get 99% accuracy with the Windows Vista Speech Recognition system, every day. I know a lot of other people here at work that use it regularly get similar results (to mine -- not Shanen's!)

and another one:
Speech demos in front of a large audience are particularly tricky because the acoustics change from when you did your rehearsal (in an empty room) to when it's showtime (lots of people). There are plenty of ways to make up for that, of course and the best way is to have a good close-talk microphone. But even then, sigh, something can happen.

No comments:

Saturday, March 15, 2008

Windows Vista speech recognition failure - the whole story



Here is what microsoft team is saying about this:

This specific problem looks like it was likely caused by a serious audio gain bug in the specific build that Shanen was using for the FAM (Financial Analysts Meeting) demo. I'm actually going to be taking a look at the machine personally this morning to see if that's what it was or if it was something else. No matter what it was, we'll address it before we ship in one way or another.

The results Shanen got during the demo were not at all similar to the results he was getting earlier yesterday morning, so there's certainly something else going on. If it always performed this poorly, we wouldn't ship it! :-)

I consistently get 99% accuracy with the Windows Vista Speech Recognition system, every day. I know a lot of other people here at work that use it regularly get similar results (to mine -- not Shanen's!)

and another one:
Speech demos in front of a large audience are particularly tricky because the acoustics change from when you did your rehearsal (in an empty room) to when it's showtime (lots of people). There are plenty of ways to make up for that, of course and the best way is to have a good close-talk microphone. But even then, sigh, something can happen.

No comments: