Java FrontEnd类代码示例

import edu.cmu.sphinx.frontend.FrontEnd; //导入依赖的package包/类
/**
 * Creates a new voice activity detector
 * @param ais input stream like LocalMicrophone or SocketMicrophone
 * @param AisName name of the microphone LocalMicrophone or SocketMicrophone
 */
@SuppressWarnings({ "rawtypes", "unchecked" })
public VoiceActivityDetector(AudioInputStream ais, String AisName) {
	super(null,new AudioFormat(16000, 16,	1, true, false), 0);


	//audio source with 3200 byte read per read
	AudioFileDataSource audioDataSource = new AudioFileDataSource(3200,
			null);
	audioDataSource.setInputStream(ais, AisName);
	
	ArrayList pipeline = new ArrayList();
	pipeline.add(audioDataSource);
	//blocks data into frames
	pipeline.add(new DataBlocker(10));
	//classifies speech frames
	pipeline.add(new SpeechClassifier(10, 0.015, 10, 0));
	//marks as speech
	pipeline.add(new SpeechMarker(200, 300, 100, 30, 100, 15.0));
	//removes non speech
	pipeline.add(new NonSpeechDataFilter());

	frontend = new FrontEnd(pipeline);

}
 

import edu.cmu.sphinx.frontend.FrontEnd; //导入依赖的package包/类
/**
 * The learning rate determines how a newly trained model is merged
 * with the existing one according to:
 *
 * 	 model_values = lambda * new_model_values + (1-lambda) * old_model_values
 *
 * That means if the learning rate is 0, no new values are applied and
 * the model is unchanged. If the learning rate is 1, the new model replaces
 * the old model. Intermediate values merge the values of both models.
 *
 * To disable training, set the learning rate to 0.
 *
 * @param src
 * @param learningRate
 */
public VQVADPipeline(AudioFileDataSource src, double learning_rate) {
	ArrayList<DataProcessor> pipeline = new ArrayList<DataProcessor>();

	float frame_length_ms = 30;
	float frame_shift_ms = 10;
	double lower_freq = 0;

	trainer = new VQVADTrainer();

	pipeline.add(src);
	pipeline.add(new Dither());
	pipeline.add(new RaisedCosineWindower(0, frame_length_ms, frame_shift_ms));
	pipeline.add(new MFCCPipeline(lower_freq, src.getSampleRate()/2, 27));
	pipeline.add(trainer);
	pipeline.add(new VQVADClassifier(learning_rate));
	pipeline.add(new FrameOverlapFilter(frame_length_ms, frame_shift_ms));
	pipeline.add(new GapSmoothing(12));

	frontend = new FrontEnd(pipeline);
}
 

import edu.cmu.sphinx.frontend.FrontEnd; //导入依赖的package包/类
static private void processFile(String inputFile, String outputFile,
        ConfigurationManager cm) throws MalformedURLException, IOException {

    FrontEnd frontend = (FrontEnd) cm.lookup("endpointer");

    AudioFileDataSource dataSource = (AudioFileDataSource) cm
            .lookup("audioFileDataSource");
    System.out.println(inputFile);
    dataSource.setAudioFile(new File(inputFile), null);
    WavWriter wavWriter = (WavWriter) cm.lookup("wavWriter");
    wavWriter.setOutFilePattern(outputFile);

    frontend.initialize();

    Data data = null;
    do {
        data = frontend.getData();
    } while (data != null);
}
 

import edu.cmu.sphinx.frontend.FrontEnd; //导入依赖的package包/类
protected FrontEnd setupFrontend(AudioFileDataSource audioDataSource) {
	ArrayList<DataProcessor> pipeline = new ArrayList<DataProcessor>();
	// VQVAD pipeline
	pipeline.add(new VQVADPipeline(audioDataSource));

	// Mark speech start/end
	pipeline.add(new SpeechMarker(200, 400, 100, 30, 100, 15.0));

	return new FrontEnd(pipeline);
}
 

import edu.cmu.sphinx.frontend.FrontEnd; //导入依赖的package包/类
/**
	 * @param args
	 * @throws IOException
	 */
	public static void main(String[] args) throws IOException {

//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/foo.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noise_only.wav");

//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/clean/sp12.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/clean/sp03.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/clean/sp05.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/clean/sp06.wav");
		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/clean/sp07.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/train_10dB/sp07_train_sn10.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/train_5dB/sp07_train_sn5.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/sp07_train_sn10_long.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/train_10dB/sp12_train_sn10.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/train_5dB/sp12_train_sn5.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/car_10dB/sp12_car_sn10.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/sp12_train_sn10_then_noise_only.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/sp12_train_sn10_preceeding_noise.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/sp12_train_sn10_preceeding_noise_sp12_train_sn10_again.wav");


		AudioFileDataSource audioDataSource = new AudioFileDataSource(3200, null);
		audioDataSource.setAudioFile(path, "in");
		ArrayList<DataProcessor> pipeline = new ArrayList<DataProcessor>();
		pipeline.add(new VQVADPipeline(audioDataSource));
		pipeline.add(new ClassificationResultDumper(path.getFile(), 30, 10, "/tmp/vqvad_classification_result"));

		FrontEnd frontend = new FrontEnd(pipeline);
		Data d;

		do {
			d = frontend.getData();
		} while(d != null && !(d instanceof DataEndSignal));
	}
 

import edu.cmu.sphinx.frontend.FrontEnd; //导入依赖的package包/类
public static void main(String[] args) throws MalformedURLException {
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/foo.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset//noise_only.wav");

//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/clean/sp12.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/clean/sp06.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/clean/sp07.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/train_10dB/sp07_train_sn10.wav");
		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/train_5dB/sp07_train_sn5.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/train_10dB/sp12_train_sn10.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/train_5dB/sp12_train_sn5.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/car_10dB/sp12_car_sn10.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/sp12_train_sn10_then_noise_only.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/sp12_train_sn10_preceeding_noise.wav");
//		URL path = new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/sp12_train_sn10_preceeding_noise_sp12_train_sn10_again.wav");


		AudioFileDataSource audioDataSource = new AudioFileDataSource(3200, null);
		audioDataSource.setAudioFile(path, "in");

		ArrayList<DataProcessor> pipeline = new ArrayList<DataProcessor>();
		pipeline.add(audioDataSource);
		//blocks data into frames
		pipeline.add(new DataBlocker(10));
		//classifies speech frames
		pipeline.add(new SpeechClassifier(10, 0.015, 10, 0));
		pipeline.add(new GapSmoothing(12));
		pipeline.add(new ClassificationResultDumper(path.getFile(), 10, 10, "/tmp/vqvad_classification_result"));

		FrontEnd frontend = new FrontEnd(pipeline);
		Data d;

		do {
			d = frontend.getData();
		} while(d != null && !(d instanceof DataEndSignal));
	}
 

import edu.cmu.sphinx.frontend.FrontEnd; //导入依赖的package包/类
public MFCCPipeline(double minFreq, double maxFreq, int numFilters) {
	final ArrayList<DataProcessor> pipeline = new ArrayList<DataProcessor>();

	singleDataBuffer = new SingleDataBuffer();
	denoisedFrameBuffer = new SingleDataBuffer();

	pipeline.add(singleDataBuffer);
	pipeline.add(new DiscreteFourierTransform());
	pipeline.add(new Denoise(0.7, 0.995, 0.5));
	pipeline.add(denoisedFrameBuffer);
	pipeline.add(new MelFrequencyFilterBank(minFreq, maxFreq, numFilters));
	pipeline.add(new DiscreteCosineTransform2(numFilters, 12));

	frontend = new FrontEnd(pipeline);
}
 

import edu.cmu.sphinx.frontend.FrontEnd; //导入依赖的package包/类
/** Returns a the next <code>DataProcessor</code> of type <code>predecClass</code> which precedes <code>dp</code>
 * @param dp
 * @param predecClass
 */
public static <T extends DataProcessor> T getFrontEndProcessor(DataProcessor dp, Class<T> predecClass) {
    while (!predecClass.isInstance(dp)) {
        if (dp instanceof FrontEnd)
            dp = ((FrontEnd) dp).getLastDataProcessor();
        else
            dp = dp.getPredecessor();

        if (dp == null)
            return null;
    }


    return predecClass.cast(dp);
}
 

import edu.cmu.sphinx.frontend.FrontEnd; //导入依赖的package包/类
/**
 * Creates a new SpectrogramPanel for the given AudioData.
 *
 * @param frontEnd   the front end to use
 * @param dataSource the source of audio
 * @param audioData  the AudioData
 */
public SpectrogramPanel(FrontEnd frontEnd,
                        StreamDataSource dataSource, AudioData audioData) {
    audio = audioData;
    this.frontEnd = frontEnd;
    this.dataSource = dataSource;
    audio.addChangeListener(new ChangeListener() {
        @Override
        public void stateChanged(ChangeEvent event) {
            computeSpectrogram();
        }
    });
}
 

import edu.cmu.sphinx.frontend.FrontEnd; //导入依赖的package包/类
/**
 * @param args
 * @throws IOException
 */
public static void main(String[] args) throws IOException {

	URL[] paths = new URL[]{
		new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/clean/sp12.wav"),
		new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/clean/sp03.wav"),
		new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/clean/sp05.wav"),
		new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/clean/sp06.wav"),
		new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/clean/sp07.wav"),
		new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/train_10dB/sp05_train_sn10.wav"),
		new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/train_10dB/sp06_train_sn10.wav"),
		new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/train_10dB/sp07_train_sn10.wav"),
		new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/train_5dB/sp07_train_sn5.wav"),
		new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/sp07_train_sn10_long.wav"),
		new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/train_10dB/sp12_train_sn10.wav"),
		new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/train_5dB/sp12_train_sn5.wav"),
		new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/noizeus_train/car_10dB/sp12_car_sn10.wav"),
		new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/sp12_train_sn10_then_noise_only.wav"),
		new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/sp12_train_sn10_preceeding_noise.wav"),
		new URL("file:///home/nemo/Documents/Studium/Master/study/code/VQVAD/trainset/sp12_train_sn10_preceeding_noise_sp12_train_sn10_again.wav"),
	};

	VQVADPipeline vadpipe = null;
	VQVADModel lastModel = null;

	for (URL path : paths) {
		System.out.println(path);
		AudioFileDataSource audioDataSource = new AudioFileDataSource(3200, null);
		audioDataSource.setAudioFile(path, "in");

		vadpipe = new VQVADPipeline(audioDataSource);

		if (lastModel != null) {
			vadpipe.setStartingModel(lastModel);
		}

		ArrayList<DataProcessor> pipeline = new ArrayList<DataProcessor>();
		pipeline.add(vadpipe);
		pipeline.add(new ClassificationResultDumper(path.getFile(), 30, 10, "/tmp/vqvad_classification_result"));

		FrontEnd frontend = new FrontEnd(pipeline);
		Data d;

		do {
			d = frontend.getData();
		} while(d != null && !(d instanceof DataEndSignal));

		lastModel = vadpipe.getCurrentModel();
	}
}