Long audio speech recognition on Android

问题

I want to develop a module which will use a speech to text support in Android. I found out many documentation and demos related to RecognizerIntent and such others. But I found that all of such demos just fetch the voice till 10 secs or so. But I want my demo to run for more than 5-10 minutes. I don't have any issue if that is not running offline, as my app is always working online.

I have also looked in to Pocketsphinx on Android, but that didn't worked out well. Also, that gave support just for Android Studio and not on Eclipse.

I have seen many apps giving the facility to convert speech to text for 5-10 mins continuously, such as: Speech To Text Notepad.

Can anyone suggest any other library of demo code with which this can be achieved ? TIA.

回答1:

Please refer this Android Speech Recognition Without Dialog In A Custom Activity.

try override method onEndOfSpeech and restart service again using speechRecognizer.startListening(recognizerIntent)

I got the same result as the app you referenced Speech To Text Notepad, here is my activity

import java.util.ArrayList;
import android.speech.RecognitionListener;
import android.speech.RecognizerIntent;
import android.speech.SpeechRecognizer;
import android.app.Activity;
import android.content.Intent;
import android.os.Bundle;
import android.util.Log;
import android.view.View;
import android.view.WindowManager;
import android.widget.CompoundButton;
import android.widget.CompoundButton.OnCheckedChangeListener;
import android.widget.ProgressBar;
import android.widget.TextView;
import android.widget.ToggleButton;

public class VoiceRecognitionActivity extends Activity implements
        RecognitionListener {

    private TextView returnedText;
    private ToggleButton toggleButton;
    private ProgressBar progressBar;
    private SpeechRecognizer speech = null;
    private Intent recognizerIntent;
    private String LOG_TAG = "VoiceRecognition";
    String speechString = "";
    boolean spechStarted = false;

    @Override
    protected void onCreate(Bundle savedInstanceState) {
        super.onCreate(savedInstanceState);
        setContentView(R.layout.activity_voice_recognition);
        getWindow().addFlags(WindowManager.LayoutParams.FLAG_KEEP_SCREEN_ON);
        returnedText = (TextView) findViewById(R.id.textView1);
        progressBar = (ProgressBar) findViewById(R.id.progressBar1);
        toggleButton = (ToggleButton) findViewById(R.id.toggleButton1);

        progressBar.setVisibility(View.INVISIBLE);
        speech = SpeechRecognizer.createSpeechRecognizer(this);
        speech.setRecognitionListener(this);
        recognizerIntent = new Intent(RecognizerIntent.ACTION_RECOGNIZE_SPEECH);
        recognizerIntent.putExtra(RecognizerIntent.EXTRA_LANGUAGE_PREFERENCE,
                "en");
        recognizerIntent.putExtra(RecognizerIntent.EXTRA_CALLING_PACKAGE,
                this.getPackageName());
        recognizerIntent.putExtra(RecognizerIntent.EXTRA_LANGUAGE_MODEL,
                RecognizerIntent.LANGUAGE_MODEL_WEB_SEARCH);

        recognizerIntent.putExtra(RecognizerIntent.EXTRA_PARTIAL_RESULTS,
                true);

        toggleButton.setOnCheckedChangeListener(new OnCheckedChangeListener() {

            @Override
            public void onCheckedChanged(CompoundButton buttonView,
                                         boolean isChecked) {
                if (isChecked) {
                    speech.setRecognitionListener(VoiceRecognitionActivity.this);
                    progressBar.setVisibility(View.VISIBLE);
                    progressBar.setIndeterminate(true);
                    speech.startListening(recognizerIntent);
                } else {
                    progressBar.setIndeterminate(false);
                    progressBar.setVisibility(View.INVISIBLE);
                    speech.stopListening();
                    speech.destroy();

                }
            }
        });

    }

    @Override
    protected void onPause() {
        super.onPause();
        if (speech != null) {
            speech.destroy();
            Log.i(LOG_TAG, "destroy");
        }

    }

    @Override
    public void onBeginningOfSpeech() {
        Log.i(LOG_TAG, "onBeginningOfSpeech");
        spechStarted = true;
        progressBar.setIndeterminate(false);
        progressBar.setMax(10);
    }

    @Override
    public void onBufferReceived(byte[] buffer) {
        Log.i(LOG_TAG, "onBufferReceived: " + buffer);
    }

    @Override
    public void onEndOfSpeech() {

        spechStarted = false;
        Log.i(LOG_TAG, "onEndOfSpeech");
        speech.startListening(recognizerIntent);

    }

    @Override
    public void onError(int errorCode) {
        Log.d(LOG_TAG, "FAILED ");
        if (!spechStarted)
            speech.startListening(recognizerIntent);

    }

    @Override
    public void onEvent(int arg0, Bundle arg1) {
        Log.i(LOG_TAG, "onEvent");
    }

    @Override
    public void onPartialResults(Bundle arg0) {
        Log.i(LOG_TAG, "onPartialResults");

        ArrayList<String> matches = arg0
                .getStringArrayList(SpeechRecognizer.RESULTS_RECOGNITION);

        returnedText.setText(speechString + matches.get(0));


    }

    @Override
    public void onReadyForSpeech(Bundle arg0) {
        Log.i(LOG_TAG, "onReadyForSpeech");
    }

    @Override
    public void onResults(Bundle results) {
        Log.i(LOG_TAG, "onResults");
        ArrayList<String> matches = results
                .getStringArrayList(SpeechRecognizer.RESULTS_RECOGNITION);
        speechString = speechString + ". " + matches.get(0);
    }

    @Override
    public void onRmsChanged(float rmsdB) {
        Log.i(LOG_TAG, "onRmsChanged: " + rmsdB);
        progressBar.setProgress((int) rmsdB);
    }


}

回答2:

In general, long audio speech recognition is a challenging problem, so you barely find anything open for that. Instead, I'd suggest you to apply one of the audio segmentation algorithms and recognize them separately. Also, if you have text transcript along with your audio and just want to obtain the time frames (e.g. for the video captioning problem), then the task becomes much easier, and you can try long audio alignment for that.

回答3:

I have successfully accomplished this with the help of Google Cloud Speech API. They have also added a demo here.

Google Cloud Speech-to-Text enables developers to convert audio to text by applying powerful neural network models in an easy to use API. The API recognizes 120 languages and variants, to support your global user base. You can enable voice command-and-control, transcribe audio from call centers, and more. It can process real-time streaming or pre-recorded audio, using Google’s machine learning technology.

You can transcribe the text of users dictating to an application’s microphone, enable command-and-control through voice, or transcribe audio files, among many other use cases. Recognize audio uploaded in the request, and integrate with your audio storage on Google Cloud Storage, by using the same technology Google uses to power its own products.

来源：https://stackoverflow.com/questions/33343942/long-audio-speech-recognition-on-android