Keyword or keyphrase spotting with Sphinx4

问题

I am currently trying to make my java code (using eclipse) perform some function if a certain thing is said. I am using the Sphinx4 libraries and this is what I currently have:

What I would like it to do is at the line where it says:

IF (TRUE) someFunction();

is to run the function if my speech is Hello Computer, Hello Jarvis, Good Morning Computer, or Good Morning Jarvis. Or in other words, run the function if the speech matches the "public < greet >" line of code in the .gram file. Even more specific, return "greet" if my speech corresponds with that grammar rule. I am sorry if this doesnt make sense...

Here is my listener.java file:

package speechRecognition;

import java.io.File;
import javax.sound.sampled.AudioInputStream;
import javax.sound.sampled.AudioSystem;
import javax.sound.sampled.Clip;
import edu.cmu.sphinx.frontend.util.Microphone;
import edu.cmu.sphinx.recognizer.Recognizer;
import edu.cmu.sphinx.result.Result;
import edu.cmu.sphinx.util.props.ConfigurationManager;

public class Listener {

    public void someFunction(){
        System.out.println("Did Something");
    }

    public static void main(String[] args) {
        ConfigurationManager cm;
        if (args.length > 0) { cm = new ConfigurationManager(args[0]);
        } else { cm = new ConfigurationManager(Listener.class.getResource("configurations.config.xml")); }

        Recognizer recognizer = (Recognizer) cm.lookup("recognizer");
        recognizer.allocate();

        Microphone microphone = (Microphone) cm.lookup("microphone");
        if (!microphone.startRecording()) {
            System.out.println("Cannot start microphone.");
            recognizer.deallocate();
            System.exit(1);
        }

        while (true) {
            Result result = recognizer.recognize();
            if (result != null) {
                String resultText = result.getBestFinalResultNoFiller();
                if (resultText != "" && resultText != null) {
                    IF (TRUE) someFunction();
                }
            } else {
                System.out.println("I can't hear what you said.\n");
            }
        }
    }
}

And here is my dictionary.gram:

#JSGF V1.0;
grammar dictionary;

public <greet> = (Hello | Good Morning) (Jarvis | Computer);

回答1:

You can do that but the only thing is that you need to enable OOG spotting in sphinx4. Basically take any sphinx4 grammar example and enable OOG according to this wiki page in config file:

<component name="flatLinguist"
           type="edu.cmu.sphinx.linguist.flat.FlatLinguist">
     ....
    <property name="addOutOfGrammarBranch" value="true"/>
    <property name="outOfGrammarProbability" value="1E-20"/>
    <property name="phoneInsertionProbability" value="1E-10"/>
    <property name="phoneLoopAcousticModel" value="wsj"/>
     ...
</component>

After that it will return you <unk> word as a result if just random word is recorded and will return you a key phrase if the word from your grammar is recorded.

You need to tune outOfGrammar probability to get a reliable detection. For more details see

http://cmusphinx.sourceforge.net/wiki/sphinx4:rejectionhandling

来源：https://stackoverflow.com/questions/16994980/keyword-or-keyphrase-spotting-with-sphinx4

标签

java

grammar

speech-recognition

cmusphinx

sphinx4