Search code examples
javaspeech-recognitionvoice-recognitionspeech-to-textcmusphinx

How to detect out of vocabulary word in CMU Sphinx


I am having a problem with the Sphinx voice recognition library for Java. I am using it to get input and handle it. In the grammar file , I wrote like this:

#JSGF V1.0;

grammar hello;

public <sentence> = (play | pause | next | previous);

My grammar is simple , just includes 4 words : "play" , "pause" , "next" , "previous". I have used Sphinx to detect them sucessfully . But I want my app to show a message like : "Unrecognized word" when I speak some words that do not belong to the grammar. Currently, For example, if I speak to the microphone a not belong to the grammar like :"stop" , it still show up the word that it detects that it is the nearest result.

My code is like this :

public class SphinxDemo {

    static int i = 1;
    static String resultText;

    public static void main(String[] args) {
        try {
            URL url;
            if (args.length > 0) {
                url = new File(args[0]).toURI().toURL();
            } else {
                url = SphinxDemo.class.getResource("helloworld.config.xml");
            }

            System.out.println("Loading...");

            ConfigurationManager cm = new ConfigurationManager(url);

            Recognizer recognizer = (Recognizer) cm.lookup("recognizer");
            Microphone microphone = (Microphone) cm.lookup("microphone");

            /* allocate the resource necessary for the recognizer */
            recognizer.allocate();

            /* the microphone will keep recording until the program exits */

            if (microphone.startRecording()) {
                System.out
                        .println("Say: play|pause|previous|next");

                while (true) {

                    System.out
                            .println("Start speaking. Press Ctrl-C to quit.\n");

                    Result result = recognizer.recognize();
                    if (result != null) {

                        System.out.println("Enter your choise" + "\n");
                        resultText = result.getBestFinalResultNoFiller();
                        System.out.println("You said: " + resultText + "\n");
                    }

                    if(!(resultText.equalsIgnoreCase("play") || resultText.equalsIgnoreCase("previous") || resultText.equalsIgnoreCase("pause")||resultText.equalsIgnoreCase("next"))){
                        System.out.println("Unrecognized word\n");
                    }

                }
            } else {
                System.out.println("Cannot start microphone.");
                recognizer.deallocate();
                System.exit(1);
            }

        } catch (IOException e) {
            System.err.println("Problem when loading SphinxDemo: " + e);
            e.printStackTrace();
        } catch (PropertyException e) {
            System.err.println("Problem configuring SphinxDemo: " + e);
            e.printStackTrace();
        } catch (InstantiationException e) {
            System.err.println("Problem creating SphinxDemo: " + e);
            e.printStackTrace();
        }

    }
}

I have tried to add something like this to detect unrecognized word but it does not work:

  if(!(resultText.equalsIgnoreCase("play") || resultText.equalsIgnoreCase("previous") || resultText.equalsIgnoreCase("pause")||resultText.equalsIgnoreCase("next"))){
                System.out.println("Unrecognized word\n");
 }

Solution

  • If you use latest cmusphinx, it will return <unk> when word is not in the grammar.