MDX Now does Speech Recognition!

**2k1Toaster** · 06-12-2008, 08:28 AM

I have just put a rough implementation of Speech Recognition into Uber MDX.

A very basic version of this was available for version 2 (which is now fully discontinued), but with Vista changing things up with the audio channels and speech recognition using SAPI 5.3 I made a whole new version.

This version is much more flexible with what it can say. For instance you can declare custom queries with beginnings, middles, and ends. The program will then confobulate (

) all the possible combinations into commands.

So if it was:
Beginning: "he", "she", "it"
Middle: "ran", "walked"
Ending: "into a door", "in a marathon"

All of the commands would be:
"he ran into a door", "he ran in a marathon", "he walked into a door", "he walked in a marathon", "she ran into a door", "she ran in a marathon", "she walked into a door", "she walked in a marathon", "it ran into a door", "it ran in a marathon", "it walked into a door", "it walked in a marathon".

Now with 3/2/2 possibilities, it isnt too big of a deal. But with an infinate number available, you can create commands that will respond to nearly anything effortlessly.

Also you can define what the global system variables are too for inputs, outputs, and even assign custom names to variables and digital outputs. So instead of "turn on digital output zero", you can say "initialize water cannon!" and same thing can happen, or "stand down red alert".

I have been having some fun with it and so far it works very well. I am using a cheapo mic from a Dell we got in 2000 into my laptop with Sirius radio blaring pretty loudly in my room and it can tell when I am speaking 98% of the time.

There is also a speech debugger that will show all possible commands as MDX has interpretted the configuration file, as well as display a message and the "heard" sentence/words for every hypothesis, recognition, and rejection. Sort of interesting to see.

I have not tested this with XP yet, but it should work...

The update should be ready for you guys tonight.

Any suggestions on what you want control over via voice?

Also being the Canadian I am, I programmed in some politeness which can be turned off as well. I still say thankyou to my GPS lady when she says "In 200 feet turn left onto blah blah street", so if you say thankyou to MDX it will say "you are welcome" but like I said you guys can turn it off if you want.

**2k1Toaster** · 06-13-2008, 02:44 AM

Ok, the update is out. You should get the message to update soon, or right click the MDX Updater icon in the system tray, and choose Check for Updates.

It will download and replace your main exe file. Also it will create the file FusionConfigurationSPEECH.xml. This new file is simple my example configuration file. It is not editable in the configurator, so no dont try it. Imade it a new file so it doesnt overwrite all the custom stuff you have already placed in it. Inside you will notice the speech section is now a main header, as well as the extra debugging speech window in the general section.

Give it a whirl and let me know how it goes.

**RonS** · 09-01-2008, 11:02 AM

Hi... What is required for speech? My mic is on the way but I would like to have everything ready when it gets here. Also does anyone have a good example for using speech control.

Ron

**2k1Toaster** · 09-01-2008, 02:53 PM

Speech needs a mic, .NET 3.5, and as of now, Vista. I have a XP compatible speech build, but there are significant changes to the underlying structure of it, and it is not quite ready to be public.

For some people speech recognition wont load on XP due to the way XP requires the speech thread to be initialized.

As far as examples, if you look in MDX directory, there should be a file called FusionConfigurationSPEECH.xml with a good example in there. Rename without the "SPEECH" and MDX will load it and you will see.

**RonS** · 09-01-2008, 03:25 PM

The file is not in there. But I may have removed it. I will install on another drive and see. I hope the XP version is out soon. I have tiny xp running and also have tiny vista. But tiny vista did not want to work with FB. And as I am going all solid state drives on the computer I will not have room for full vista. Keep up the good work.

**2k1Toaster** · 09-01-2008, 03:37 PM

Well you may have to manually install dependencies if you want tiny anything after the installer. It will install .NET 2.0, .NET 3.0, and .NET 3.5 full versions to account for any future dependencies.

If you already have 2.0 installed, the only thing you need from 3.5 now is the speech support for SAPI 5.

**RonS** · 09-01-2008, 06:05 PM

I reinstalled Tiny Vista. And installed MDX. When I try to run FCC I get the following error. dw20.exe The instruction at 0x70b727e7 referenced memory at 0x00000000. The memory could not be read. I cannot install .net 2.0 because it says it is part of 3.5. And the installer does not install it. The configurator does work with no problems. Any ideas?

**2k1Toaster** · 09-01-2008, 07:21 PM

if the configurator works, then 2.0 is installed.

dw20 is part of Office... Not sure how that got there. I dont have any references to Microsoft Office at all, nor do I use a doc/xls/ect.. format on it.

What version of Windows Installer do you have? You need 3.1 to install 3.5 properly.

I do not know much about TinyVista at all, so I am not going to be much help.

Does this tinyVista install the SAPI voices? I dont include them because they are HUGE.

**RonS** · 09-01-2008, 09:00 PM

No it does not install the voices. And I am not sure which installer it is. I went back to xp for now. But I think I will just install a 40 gig laptop drive and install full vista. It will be a lot easier. Then when I get the time I will play with the other and let you know how it goes. This will also let me get the atom board so I can put 2gig of memory in it.

Thanks for the help.

**Dan2008** · 09-02-2008, 06:53 AM

Originally Posted by 2k1Toaster

...dw20 is part of Office... Not sure how that got there...

It is the Windows Error Reporting tool that gathers error information for Office components and "allows" you to send error information to Microsoft.

Here is a link on how to disable it:

http://support.microsoft.com/kb/841477