Voice recognition?

[Yumda]

thanks for everyone's feedback, this has been a wonderful forum to come to. I look forward to what Litdev would come up for his next extension on this if he is able to work on it. Much appreciated

[Yumda]

Just an update of another problem. tried the sample LDspeech program on my all in ONE windows 10 pro machine and it does not work at all, tried to speak all the words listed in the array as well.  I can see that the voice is being detected by the computer's microphone ( called the microphone array - intel smart sound technology)  but it is not being pick up at all and actioned by the program.  In the sound setting, I can see that it said tmp7A0B.exe is currently using the microphone, and a tiny desktop icon on the bottom right hand corner has a speaker icon showed up.  When hover the mouse cursor over the icon it said "Small Basic Graphic window is using your microphone". Hope these information helps.



Just thought to put in what I found.

[litdev]

Thanks,

Yes, I had also come to conclusion it was not just a Win 11 issue.  I think the test I did on old win 10 laptop was using an old version of litDev extension, so there is is something in recent versions that is interfering in some way...

[Yumda]

at least we are now narrowing down the cause... Hopefully.. but still thanks for your expertise.

[litdev]

I have uploaded a version of LitDev extension where the speech recognition sample works for me ("en-GB" locale), but I would expect it to work with your OS locale as long as you have set up speech recognition for your locale and you add appropriate words to LDSpeech.Vocabulary.

This update is for the beta version only (1.2.29.8) and is avaible on my website or using extension manager in SB-Prime.

I don't understand why the changes I made work, so I will probably do a bit more testing and am grateful for any feedback.

[Scout]

Thank you very much! The example program now also works on WIN7.
Amazingly, even without an update with the version
   

By setting the culture to "en-US" I still got the exception window.
This has changed with the update to the current beta version . . .8.
The following message now appears in the text window:
LDSpeech.set__SpeechRecognition : No recognizer of the required ID found.
Parameter name: culture

[Yumda]

tried the new update on the windows 10 and even set up region and speech to various regions and the program still won't recognize the command in the sample program for LDspeech.

Regards

Quick update:

Figured out why it was not working, as English is not my native tongue, I need to tick the  " Recognize non-native accent for this language" option, and set the tolerance in the sample language to lower than 0.5.  Thought to report back in case someone is having the same problem.

[litdev]

Hi,

I have tested my current version on Win 10 and Win 11, both working for me.

Since you are still having issues I have updated the beta version (1.2.29.9) of the extension to include a DebugMode option.  This will fire the Listen event when any speech is detected, even it is not reccognised by the vocabulary.  Test program below:

Please try after setting up speech recognition on your PC in prefered language.  Do NOT try to change the locale in the SB code at the moment.

Code:

vocab[1] = "red"
LDSpeech.DebugMode = "True"
LDSpeech.Vocabulary(vocab)
LDSpeech.Listen = OnListen

While ("True")
  Program.Delay(10)
EndWhile

Sub OnListen
  TextWindow.WriteLine(LDSpeech.LastSpoken)
EndSub

[Yumda]

So, I tried the new debug code, it is now working at detecting the voice. An interesting observation, I noticed it will keep showing RED even though I tried to say other colours like blue or green. Not sure if it is the sensitivity or my accent? I modified the code by using the lastspokenconfidence and it is all up at the 0.8 range.

Any ideas?

Regards

[litdev]

Hi,
Glad we have something working.  If you set a vocabulary, it can only recognise words in the provided list.  If you don't set a vocaulary, then it uses an internal dictionary, which I have found doesn't work well.  In the debug test example I only added one word, so that is all it can detect and maybe you need to do more voice recognition training on your PC and try adding more words to the vocabulary list to get fewer false positives.

Now something is working you can experiment a bit.  

You mention 'there is a speech check box I need to tick as well as I need to set the tolerance to less than 0.5 in your sample program' in PM, maybe this will also help others.