Android Question Speech recognition engine crashes when started from timer - OK when started from UI

[rgarnett1955]

Hi,

I am using the speech to text based on the Vosk English model to enter commands and parameters into my App.

SpeechToText - Continuous Offline Voice Recognition - Biswajit

I have made a car "speed sentry" app using GPS speed which is completely hands free.

The VR system runs continuously so there is no button pushing prompt required.

When I activate the microphone from the UI the system works perfectly, but if I start the microphone programmatically the engine will read one utterance and then the application freezes.

The relevant code is:

'-----------------------------------------------------------------------------------------
Sub Activity_Create(FirstTime As Boolean)
    Activity.LoadLayout("Layout") ...
   
    
    vrStartTimer.Initialize("startVRTmr", 15000)
    vrStartTimer.Enabled = True
End Sub

'-----------------------------------------------------------------------------------------
Sub Activity_PermissionResult (Permission As String, Result As Boolean)
    If Permission = Starter.rp.PERMISSION_ACCESS_FINE_LOCATION Then
        If Result Then CallSub(Starter, "StartGPS")
    End If

    If Permission = Starter.rp.PERMISSION_RECORD_AUDIO Then
        If Result Then
            ProgressDialogShow2("Extracting voice model...",False)
            If File.IsDirectory(File.DirInternal, model_folder_name) Then
                ProgressDialogHide
                #If DEBUGGING
                Log("Model found")
                #End If
                vrInit
            Else
                #If DEBUGGING
                Log("Model not found")
                #End If
                If File.Exists(File.DirInternal, model_zip_name) Then
                    #If DEBUGGING
                    Log("Model zip found. Unzipping...")
                    #End If
                    Dim ar As Archiver
                    ar.AsyncUnZip(File.DirInternal, model_zip_name,File.DirInternal,"unzip")
                Else
                    #If DEBUGGING
                    Log("Model zip not found. Copying...")
                    #End If
                    Wait For (File.CopyAsync(File.DirAssets, model_zip_name, File.DirInternal, model_zip_name)) Complete (Success As Boolean)
                    If Success Then
                        #If DEBUGGING
                        Log("Model zip unzipping...")
                        #End If
                        Dim ar As Archiver
                        ar.AsyncUnZip(File.DirInternal,model_zip_name,File.DirInternal,"unzip")
                    Else
                        Toast.TMClearAndShow("Failed to extract model", True, True)
                    End If
                End If
            End If
        Else
            Toast.TMClearAndShow("Microphone access needed", True, True)
        End If
    End If
End Sub


'-----------------------------------------------------------------------------------------
Private Sub vrInit
    VoiceRecognitionOn = False
    STT.Initialize("STT", File.DirInternal&"/"&model_folder_name)
 End Sub   

'-----------------------------------------------------------------------------------------
Private Sub startVRTmr_Tick
    vrStartTimer.Enabled = False
    lblVoiceRecognition_Click
End Sub

'-----------------------------------------------------------------------------------------
Private Sub lblVoiceRecognition_Click
    If VoiceRecognitionOn = False Then
        VoiceRecognitionOn = True
        sentenceText = ""
        STT.prepareMicrophone("")
    Else
        lblVoiceRecognition.Text = ""
        If STT.stop Then
            Toast.TMClearAndShow("Voice Recognition Off", False, True)
        Else
            Toast.TMClearAndShow("Voice Recognition Failed to Stop", True, True)
        End If
        VoiceRecognitionOn = False
    End If
    lblPartialText.Visible = False
    showPartialText = False
End Sub

If I enable the startVRTmr the system hangs; if I disable it I can start the VR from the UI "lblVoiceRecognition_Click" and it runs fine.

It does this in both release and debug modes using USB Debugging on my Pixel 6A.

I will post the app in "my creations" once I have it sorted.

Has anyone any ideas on this problem?

I have attached the project excluding the vosk model file which is ~40 MB. This can be downloaded from the alphacephei web site

VOSK Models

Accurate speech recognition for Android, iOS, Raspberry Pi and servers with Python, Java, C#, Swift and Node.

alphacephei.com

Regards
Rob

 

[rgarnett1955]

Found the problem! It was the person doing the programming.

Hard to get good help!