Click here to Skip to main content
15,919,774 members
Please Sign up or sign in to vote.
0.00/5 (No votes)
See more:
I have written a C# speech recognizer based on the Microsoft Doc here. Initialize and Manage a Speech Recognition Engine (Microsoft.Speech) | Microsoft Docs[^]

But now I am trying to improve its recognition by "training" WSR somehow to better recognize my voice, like Dragon Naturally Speaking does, or like the built-in Windows voice recognition training dialog does.

But I can't find any information anywhere on how to get the user to train my own set of words. Is this possible? It must be, I think, since the Windows OS does it. But there might not be a public API for doing it.

Can anyone shed some light on this, or give me a link to some information or code examples? Thank you

What I have tried:

I have built a working recognition app, spent hours with the Microsoft doc, and searched the HotExamples website for examples.
Posted
Updated 27-May-22 6:35am
Comments
0x01AA 27-May-22 10:49am    

1 solution

Thank you for your help. I read the article (again), and his code looks generally like mine. The most interesting thing for me was he said it was almost impossible to train words from code.
Quote:
This question is asked frequently in comments: how to train your speech recognition engine? From code, it is impossible, unfortunately.


I was hoping to train specific phrases for my grammar, but he reinforced the idea that users can train the entire Windows Speech Recognition system using the default Windows-supplied training mechanism. I will try that instead; maybe it will work well enough for my purposes.
 
Share this answer
 
Comments
[no name] 27-May-22 12:52pm    
You're making a lot of assumptions, none of which include that probably AI is playing a bigger part. You should do some tests of your key words with different people and determine the "error rate", if any. I suspect it's the "mix" of a given set of words that would be the problem, instead of obvious "commands" which are usually quite distinct in their phonetics.

This content, along with any associated source code and files, is licensed under The Code Project Open License (CPOL)



CodeProject, 20 Bay Street, 11th Floor Toronto, Ontario, Canada M5J 2N8 +1 (416) 849-8900