B4J Question B4J VOSK with Greek input/language...

[Magma]

Hi there...

I ve found here the VOSK a b4j snippet... that works:

STT - Offline Speech Recognization with Vosk (JavaObject/Inline)

Hello everybody. I made an example of how to use vosk in B4J. In this example we use vosk to listen to our microphone and play the words it understands on the screen. the example is very simple and functional (at least on my PC xD) In the example you will see two labels, the top one is what...

www.b4x.com

I am downloading greek model... but the text result returns is unreadable..

when i say "test"
in greek "τεστ"

For example returns:
"της"

but the right is to return "τεστ" or "test"

Thought was something like charset for java jdk 19

So I ve added at start:
#VirtualMachineArgs: -Dfile.encoding=UTF-8 -Dsun.stdout.encoding=UTF-8 -Dsun.stderr.encoding=UTF-8
#PackagerProperty: VMArgs = -Dfile.encoding=UTF-8 -Dsun.stdout.encoding=UTF-8 -Dsun.stderr.encoding=UTF-8

but nothing better ... then tried to add some lines at java code to convert charset - getting different result - but none is the right...

    import java.nio.charset.Charset;
    private final Charset UTF8_CHARSET = Charset.forName("UTF-8");
    private final Charset PREVIOUSCHARSET = Charset.forName("ISO-8859-1");
....
                        byte ptext[] = resultText.getBytes(PREVIOUSCHARSET);
                        String ppv = new String(ptext, UTF8_CHARSET);
                        subQvEvent("recognition_result", ppv);

... what do you think ?

 

[Magma]

Forgot to mention @Douglas Farias created the script... hope to see my question

 

[Magma]

Well, found the problem ...was java jdk 19...

going back to JAVA 11 - all works perfect !


tried the following to java inline - was someway better (with java19)- not the best:

import javax.sound.sampled.*;

public static class AudioCapture {

    private AudioFormat format;
    private float sampleRate;
    private int sampleSizeInBits;
    private int channels;
    private org.vosk.Recognizer recognizer;
  
    import java.nio.charset.Charset;
    private final Charset UTF8_CHARSET = Charset.forName("UTF-8");
    private final Charset PREVIOUSCHARSET = Charset.forName("CP1253");

    public AudioCapture(float sampleRate, int sampleSizeInBits, int channels, org.vosk.Recognizer recognizer) {
        this.sampleRate = sampleRate;
        this.sampleSizeInBits = sampleSizeInBits;
        this.channels = channels;
        this.recognizer = recognizer;
        this.format = new AudioFormat(sampleRate, sampleSizeInBits, channels, true, false);
        subQvEvent("logs", "AudioCapture initialized with sampleRate: " + sampleRate + ", sampleSizeInBits: " + sampleSizeInBits + ", channels: " + channels);
    }

    private TargetDataLine line;
    private boolean running = false;

    public void startRecording() {
        subQvEvent("logs", "Start recording...");
      
        try {
            DataLine.Info info = new DataLine.Info(TargetDataLine.class, format);
            if (!AudioSystem.isLineSupported(info)) {
                subQvEvent("logs", "Line not supported");
                return;
            } else {
                subQvEvent("logs", "Line supported");
            }

            line = (TargetDataLine) AudioSystem.getLine(info);
            line.open(format);
            line.start();
            running = true;

            new Thread(() -> {
                byte[] buffer = new byte[1024];
                while (running) {
                    int numRead = line.read(buffer, 0, buffer.length);
                    if (numRead > 0) {
                        boolean result = recognizer.acceptWaveForm(buffer, numRead);
                        String resultText;
                        if (result) {
                            resultText = recognizer.getResult();
                        } else {
                            resultText = recognizer.getPartialResult();
                        }
                        byte ptext[] = resultText.getBytes(PREVIOUSCHARSET);
                        String ppv = new String(ptext, UTF8_CHARSET);
                        subQvEvent("recognition_result", ppv);
                    }
                }
                line.drain();
                line.close();
                subQvEvent("logs", "Recording stopped.");
            }).start();
        } catch (LineUnavailableException e) {
            subQvEvent("logs", "Line unavailable: " + e.getMessage());
        }
    }

    public void stopRecording() {
        subQvEvent("logs", "Stopping recording...");
        running = false;
        if (line != null) {
            line.stop();
            line.close();
            line = null;
        }
    }

    private void subQvEvent(String eventName, Object data) {
        ba.raiseEventFromUI(null, eventName.toLowerCase() + "_qv", new Object[]{data});
    }
}

add to "previous code of Douglas" those four lines (as you can see) :
private final Charset UTF8_CHARSET = Charset.forName("UTF-8");
private final Charset PREVIOUSCHARSET = Charset.forName("CP1253");
...
byte ptext[] = resultText.getBytes(PREVIOUSCHARSET);
String ppv = new String(ptext, UTF8_CHARSET);
subQvEvent("recognition_result", ppv);


but without changing anything from the first code with java 11... works super!

 

[aeric]

inei peroho

Should be good on JDK 14.