Author Topic: Voice training  (Read 4009 times)

Rephil

  • Newbie
  • *
  • Posts: 19
Voice training
« on: December 11, 2018, 11:36:53 AM »
I have run the Windows voice training several time and still voice attack will not recognize words like HOOK how can I train VA file are all showing what it sees when I say hook

Gary

  • Administrator
  • Hero Member
  • *****
  • Posts: 2832
Re: Voice training
« Reply #1 on: December 11, 2018, 11:48:35 AM »
Under normal circumstances, a trained speech engine should and will work great for you.  There could be other things involved, so, you might want to take a look at this thread:

https://forum.voiceattack.com/smf/index.php?topic=1635.0

A bit of a long read, but it's helped out quite a bit with identifying and correcting speech-related issues.

Hope that helps!

Exergist

  • Global Moderator
  • Sr. Member
  • *****
  • Posts: 405
  • Ride the lightning
Re: Voice training
« Reply #2 on: December 13, 2018, 10:58:25 AM »
Assuming you're using the built-in "System Default" speech engine (you can check via VoiceAttack Options => Recognition => Speech Engine), one thing you can do to rule out VA as the problem source (though it likely isn't) is to test out dictation directly through Windows Speech Recognition (WSR). To do this:
  • Open a program like Notepad
  • Type "Windows Speech Recognition" into the Windows search box and open the corresponding application (another route is through "Start Speech Recognition" via the Speech Recognition interface of the Control Panel, which is also accessible through VoiceAttack Options => Recognition => Utilities => Speech Control Panel)
  • Turn WSR to "on" (i.e., start listening)
  • Place Notepad in focus and then start dictating
You should see the recognized text appear as an insert box. Then you can test out the various words or phrases that VA isn't recognizing and see if the misunderstandings are more related to the speech engine built into Windows that VA is leveraging to do its thing.

You also have the option to add words and custom pronunciations to the Windows speech dictionary. The dictionary is accessible via a right click on the WSR interface. Bear in mind that the speech engine's accuracy in translating your custom recording of a word into phonetic data (that the engine can then use for improved speech recognition as well as text-to-speech) will depend on how well the speech engine is trained to recognize your voice.

Finally when it comes to recognizing commands (whatever is in the "When I say" box) VA and the speech engine in general will have greater success recognizing a combination of words (e.g., "hook fish"). You can think of these combinations as more "unique" than single words, which helps the speech engine narrow down what you said and more likely match it to the command. Another way to think about it, there are more words that sound like "hook" than there are individual words and word combinations that sound like "hook fish." Again though, most people shouldn't have this issue out of the box after proper voice training.

@Gary please correct me if I've misrepresented anything above.

HTH