Android RecognitionListener: onResults being called twice

Question

I have a project using RecognitionListener written in Kotlin. The speech-to-text function was always a success and never presented any problems.

Since last week, it's onResult function started to be called twice. No changes were made on the project. I tested old versions of the project (from months ago) and those had the same problem.

There are three different cases:

1. Small text (1 to 8 words) and SpeechRecognizer being stopped automatically -> onResult() called twice;
2. Big text (9 words or more) and SpeechRecognizer being stopped automatically -> Normal behavior (onResult() called once);
3. Any text size and SpeechRecognizer stopListening() function called manually (from code) -> Normal behavior.

Here is the VoiceRecognition speech-to-text class code:

class VoiceRecognition(private val activity: Activity, language: String = "pt_BR") : RecognitionListener {

    private val AudioLogTag = "AudioInput"

    var voiceRecognitionIntentHandler: VoiceRecognitionIntentHandler? = null
    var voiceRecognitionOnResultListener: VoiceRecognitionOnResultListener? = null //Must have this
    var voiceRecognitionLayoutChanger: VoiceRecognitionLayoutChanger? = null

    var isListening = false

    private val intent: Intent
    private var speech: SpeechRecognizer = SpeechRecognizer.createSpeechRecognizer(activity)

    init {
        speech.setRecognitionListener(this)

        intent = Intent(RecognizerIntent.ACTION_RECOGNIZE_SPEECH)
        intent.putExtra(
            RecognizerIntent.EXTRA_LANGUAGE_MODEL,
            RecognizerIntent.LANGUAGE_MODEL_FREE_FORM
        )
        intent.putExtra(RecognizerIntent.EXTRA_LANGUAGE, language)
    }

    //It is important to put this function inside a clickListener
    fun listen(): Boolean {
        if (ContextCompat.checkSelfPermission(activity, Manifest.permission.RECORD_AUDIO) != PackageManager.PERMISSION_GRANTED) {
            ActivityCompat.requestPermissions(activity, arrayOf(Manifest.permission.RECORD_AUDIO), 1)
            return false
        }

        speech.startListening(intent)

        Log.i(AudioLogTag, "startListening")

        return true
    }

    //Use this if you want to stop listening but still get recognition results
    fun endListening(){
        Log.i(AudioLogTag, "stopListening")

        speech.stopListening()
        isListening = false
    }

    fun cancelListening(){
        Log.i(AudioLogTag, "cancelListening")

        speech.cancel()
        voiceRecognitionLayoutChanger?.endListeningChangeLayout()
        isListening = false
    }

    override fun onReadyForSpeech(p0: Bundle?) {
        Log.i(AudioLogTag, "onReadyForSpeech")

        voiceRecognitionLayoutChanger?.startListeningChangeLayout()
        isListening = true
    }

    override fun onRmsChanged(p0: Float) {
//        Log.i(AudioLogTag, "onRmsChanged: $p0")
//        progressBar.setProgress((Int) p0)
    }

    override fun onBufferReceived(p0: ByteArray?) {
        Log.i(AudioLogTag, "onBufferReceived: $p0")
    }

    override fun onPartialResults(p0: Bundle?) {
        Log.i(AudioLogTag, "onPartialResults")
    }

    override fun onEvent(p0: Int, p1: Bundle?) {
        Log.i(AudioLogTag, "onEvent")
    }

    override fun onBeginningOfSpeech() {
        Log.i(AudioLogTag, "onBeginningOfSpeech")
    }

    override fun onEndOfSpeech() {
        Log.i(AudioLogTag, "onEndOfSpeech")

        voiceRecognitionLayoutChanger?.endListeningChangeLayout()
        isListening = false
    }

    override fun onError(p0: Int) {
        speech.cancel()
        val errorMessage = getErrorText(p0)
        Log.d(AudioLogTag, "FAILED: $errorMessage")
        voiceRecognitionLayoutChanger?.endListeningChangeLayout()
        isListening = false
    }

    override fun onResults(p0: Bundle?) {

        val results: ArrayList<String> = p0?.getStringArrayList(SpeechRecognizer.RESULTS_RECOGNITION) as ArrayList<String>

        Log.i(AudioLogTag, "onResults -> ${results.size}")

        val voiceIntent: Int? = voiceRecognitionIntentHandler?.getIntent(results[0])
        if (voiceIntent != null && voiceIntent != 0) {
            voiceRecognitionIntentHandler?.handle(voiceIntent)
            return
        }

        voiceRecognitionOnResultListener!!.onResult(results[0])
    }

    private fun getErrorText(errorCode: Int): String {
        val message: String
        when (errorCode) {
            SpeechRecognizer.ERROR_AUDIO -> message = "Audio recording error"
            SpeechRecognizer.ERROR_CLIENT -> message = "Client side error"
            SpeechRecognizer.ERROR_INSUFFICIENT_PERMISSIONS -> message = "Insufficient permissions"
            SpeechRecognizer.ERROR_NETWORK -> message = "Network error"
            SpeechRecognizer.ERROR_NETWORK_TIMEOUT -> message = "Network timeout"
            SpeechRecognizer.ERROR_NO_MATCH -> message = "No match"
            SpeechRecognizer.ERROR_RECOGNIZER_BUSY -> message = "RecognitionService busy"
            SpeechRecognizer.ERROR_SERVER -> message = "Error from server"
            SpeechRecognizer.ERROR_SPEECH_TIMEOUT -> message = "No speech input"
            else -> message = "Didn't understand, please try again."
        }
        return message
    }

    //Use it in your overriden onPause function.
    fun onPause() {
        voiceRecognitionLayoutChanger?.endListeningChangeLayout()
        isListening = false

        speech.cancel()
        Log.i(AudioLogTag, "pause")
    }

    //Use it in your overriden onDestroy function.
    fun onDestroy() {
        speech.destroy()
    }

listen(), endListening() and cancelListening() are all called from a button.

Answer 1

I found this open issue: https://issuetracker.google.com/issues/152628934

As I commented, I assume it is an issue with the "speech recognition service" and not with the Android RecognitionListener class.

Answer 2

this is my temporary workaround

singleResult=true;

@Override
 public void onResults(Bundle results) {
            Log.d(TAG, "onResults"); //$NON-NLS-1$
            if (singleResult) {
                ArrayList<String> matches = results.getStringArrayList(SpeechRecognizer.RESULTS_RECOGNITION);

                if (matches != null && matches.size() > 0) {
                    Log.d("single Result", "" + matches.get(0));

                }
                singleResult=false;
            }

            getHandler().postDelayed(new Runnable() {
                @Override
                public void run() {
                    singleResult=true;

                }
            },100);

        }

Answer 3

I had the same problem and I've just added a boolean flag in my code, but ofcourse it's a temporary solution and I don't know the source of this problem.

val recognizer = SpeechRecognizer.createSpeechRecognizer(context)
recognizer.setRecognitionListener(
    object : RecognitionListener {

        var singleResult = true

        override fun onResults(results: Bundle?) {
            if (singleResult) {
                results?.getStringArrayList(SpeechRecognizer.RESULTS_RECOGNITION).let {
                    // do something with result
                }
                // next result will be ignored
                singleResult = false
            }
        }
    }

Answer 4

This just started happening in one of my apps yesterday. I added a boolean to allow the code to execute only once, but I'd love an explanation as to why it suddenly started doing this. Any updates?