B4A Library [Chargeable] SpeechRecognizer Library

mrodriguez said:

Works perfect!!!

Click to expand...

Harris said:

Great! Do you have a code example?

This is my project:
User is "out in the field" without network coverage.
Since his hands are covered with bee wax and honey, he needs to make a notation of the conditions of the yard (voice recording without typing).
When he returns to network coverage, the file (voice recording) is processed (as you stated above) and the text record returned added to the DB.

Make Sense? Doable?

Thanks

Click to expand...

mrodriguez said:

Hi, I need to use google voice recognition over rikomagic.
Rikomagic use android 4.1 but doesnt have the voice recognition software and can not install it from play store.

So my only way is to use the unofficial voice recognition api from google.
https://gist.github.com/alotaiba/1730160

The proccess is ver simplye:
I send a post to https://www.google.com/speech-api/v1/recognize with the recorded file (using audiorecorder b4a library), and the respose is an json with the text.

Question:
Are you using that for your SpeechRecognizer Library?
If not, are you using the android voice recognition software?

Thanks

Click to expand...

wimpie3 said:

How can we specify a language?

The Android docs suggest the following:

B4X:

    String languagePref = "de";//OR, whatever iso code...
    Intent.putExtra(RecognizerIntent.EXTRA_LANGUAGE, languagePref);
    Intent.putExtra(RecognizerIntent.EXTRA_LANGUAGE_PREFERENCE, languagePref);
    Intent.putExtra(RecognizerIntent.EXTRA_ONLY_RETURN_LANGUAGE_PREFERENCE, languagePref);

But how do we do this in your library?

Click to expand...

Rick Harris said:

I just bought this library. It works very well. Regrettably only the default language is supported. However, XverhelstX has been very helpful by providing some source code to add other languages. I have asked him to include it in his library because it is not b4A code (it uses Google API code). I don't know yet whether I will be able to implement it in order to add other languages. The library also does not have any VU audio meter code. This is really a pity because intercepting "rms" events is extremely simple. To do so, modify the provided BAS file:

B4X:

'Rms has changed
Sub SpeechRecognition_RmsChanged (rms AsFloat)
  Log("RMS: " & rms)
  CallSub2(Main, "ShowVU", rms)
End Sub

And add a routine to main:

B4X:

Sub ShowVU(rms AsFloat)
  Dim VUlevel1 as Int
  VULevel1 = rms * 10
  'etc....
End Sub

Click to expand...

Sub StartListening  
  Dim i AsIntent
  i.Initialize("android.speech.action.RECOGNIZE_SPEECH", "")
  i.PutExtra("android.speech.extra.LANGUAGE_MODEL", "free_form")
  i.PutExtra("calling_package", "your package name")
  i.PutExtra("android.speech.extra.MAX_RESULTS", 5)
  i.PutExtra("android.speech.extra.LANGUAGE", "en-US")
  sr.StartListening(sr.CreateRecognizeIntent2(i))
End Sub

brelto85 said:

The CreateRecognizeIntent2 method there isn't in SpeechRecognizer library
is a method that should be created by author or by us?
How to do this?

Click to expand...

Rick Harris said:

Unfortunately I have ran into a problem with the speechrecognizer on a HTC smartphone. Some of my customers are starting to report this problem as well. The microphone is so sensitive that background noise is easily picked up. Regrettably the AGC (Automatic Gain Control) increases the microphone's sensitivity as soon as the user has finished speeking. The recognizer believes the increasing background noise to be speech and therefore the SpeechRecognition_EndOfSpeech event is fired too late or not even at all (the recogniser then times-out). This problem can be witnessed by following the SpeechRecognition_RMSchanged event that clearly shows the microphone volume in relation to the background noise (on a Nexus-7 everything is fine).
This problem does not occur while using B4A's default VoiceRecognizer (VR). How to solve this problem?
I was thinking of:
1. Finding a way to controle the AGC by software (but that is risky because it would be device-dependent).
2. Permanently reducing the microphone's sensitivity (ditto risky)
3. Activating a low-pass audiofilter (if possible) to reduce the pickup of noise.
4. Reducing the silence listening time with:
i.PutExtra("android.speech.extras.SPEECH_INPUT_COMPLETE_SILENCE_LENGTH_MILLIS",500)
However, so far this has been unsuccesfull because silence is not properly detected.
5. Finding out how Google has overcome this problem.

Any thoughts?

Click to expand...