VC (nonSP) - SAPI

[Hiryu]

Jitterjames!  I tried what you advised, and copied my VC SP folder and cloned it, renamed it.  Then I used only VoxCommando.exe, instead of SP so that I can use the standard engine.  I'm getting way more false positives (even setting the confidence level at 85 or above), then I was with SP. 

Then I deleted that whole folder and did a completely clean install.  Same thing.  Now I've been away from VC for a long time, so am definitely rusty, and started with SP, so that's all I'm familiar with in terms of the setup process.  Are there things I'm forgetting to do in Windows to tune the voice recognition levels? 

I'm using an open mic for now (CAD u7), but again it seems SP behaves really well, but VC Standard is picking up lots of things even when I'm not speaking (a cough by someone in the room was interpreted as launch kodi on the clean install, a tap on the desk issued the command "go up", or initiated a google search on the non-clean install).

Could it be just that the SP speech engine is just better with an open mic and that VC Standard is super sensitive with an open mic setup?

[jitterjames]

Could it be just that the SP speech engine is just better with an open mic and that VC Standard is super sensitive with an open mic setup?

It is possible, although I have not personally found this to be the case, nor has anyone else mentioned this behaviour to me.  I'm not aware of any major differences in the way that the two engines process audio.  I have generally found that I get better results with the non-SP version.

If your mic is picking up too much then you probably just need to turn the input levels down (maybe way down) on the microphone.  I would never even consider using a boundary layer microphone unless I was using prefix mode.

In the end if you find that the SP version works better for you then stick with that.

If you are wondering what you might have missed you should take another look at the wiki.

[Soda97]

Yea, I'm getting a lil frustrated too going back and forth with the standard and SP version.

I'm using an open mic setup with Kinect.  To me, the standard version with a trained profile is better by yourself.  If everyone else is using it too then I think the SP version is more suitable in this case.  However, the SP version does not support the payload dictionary.  Also, it only limited to using 1 mic for me.  Not sure, is that a bug, a limitation or something wrong with my system.

[jitterjames]

Yea, I'm getting a lil frustrated too going back and forth with the standard and SP version.
  Also, it only limited to using 1 mic for me.  Not sure, is that a bug, a limitation or something wrong with my system.

This is neither a limitation nor a bug that I'm aware of.

[Soda97]

Thanks for the reply!

So I did some more research and I see that you can set your speech input.  After doing that, my problem was fixed even after I restart VC. 

I don't know why and what happen but if anyone else happen to come up with this same problem might give this a try.

"VC.SetSpeechInput"