edu.cmu.sphinx.tools.batch
Class BatchNISTRecognizer

java.lang.Object
  edu.cmu.sphinx.tools.batch.BatchModeRecognizer
      edu.cmu.sphinx.tools.batch.BatchNISTRecognizer

All Implemented Interfaces:

Configurable

Direct Known Subclasses:

BatchForcedAlignerRecognizer

public class BatchNISTRecognizer
extends BatchModeRecognizer

Copyright 1999-2002 Carnegie Mellon University. Portions Copyright 2002 Sun Microsystems, Inc. Portions Copyright 2002 Mitsubishi Electric Research Laboratories. All Rights Reserved. Use is subject to license terms.

See the file "license.terms" for information on usage and redistribution of this file, and for a DISCLAIMER OF ALL WARRANTIES.

User: Peter Wolf Date: Nov 10, 2005 Time: 2:42:06 PM Copyright 2005, Peter Wolf

Runs a NIST corpus as used by the GALE project. The inputs are a CTL file, and a REF file. The output is a CTM file.

A CTL file contains a list of utterances to decode. The format is

The is a base to which the property "dataDirectory" is prepended, and ".raw" is appended. The utterance file should be raw PCM that agrees with the "bitsPerSample", "channelCount", "samplesPerSecond", and "framesPerSecond" properties.

The and are specified in frames, where

bytesPerFrame = (bitsPerSample/8)*channelCount*samplesPerSecond/framesPerSecond

The should be a unique string. For example "__".

A REF file contains the correct transcripts of the utterances specified in the CTL file. Each line should be of the form

()

The output is a "processed" CTM file. It is used by the NIST tools to compute the performance on the copus. The format is not documented because it is currently a hack to get the Dry Run going. We need to think more about it. If you want to use this tool talk to Peter Wolf, or Arthur Chan.

Nested Class Summary

protected class	BatchNISTRecognizer.CTLException
protected class	BatchNISTRecognizer.CTLIterator
class	BatchNISTRecognizer.CTLUtterance

Field Summary

protected int	bitsPerSample
protected int	bytesPerFrame
protected int	channelCount
protected java.lang.String	ctlFile
protected java.lang.String	ctmFile
protected java.lang.String	dataDir
protected int	framesPerSecond
static java.lang.String	PROP_BITS_PER_SAMPLE The sphinx properties that specify the format of the PCM audio in the data file
static java.lang.String	PROP_CHANNEL_COUNT
static java.lang.String	PROP_CTL_FILE The property that specifies the file containing the corpus utterance audio
static java.lang.String	PROP_CTM_FILE The property that specifies the the directory where the output XXX files should be placed
static java.lang.String	PROP_DATA_DIR The property that specifies the file containing the corpus utterance audio
static java.lang.String	PROP_FRAMES_PER_SECOND
static java.lang.String	PROP_REF_FILE The property that specifies the file containing the transcripts of the corpus
static java.lang.String	PROP_SAMPLES_PER_SECOND
protected java.lang.String	refFile
protected int	samplesPerSecond

Fields inherited from class edu.cmu.sphinx.tools.batch.BatchModeRecognizer

batchManager, cm, count, curBatchItem, inputDataProcessors, logger, name, PROP_COUNT, PROP_INPUT_DATA_PROCESSORS, PROP_RECOGNIZER, PROP_SKIP, PROP_TOTAL_BATCHES, PROP_USE_POOLED_BATCH_MANAGER, PROP_WHICH_BATCH, recognizer, skip, totalBatches, usePooledBatchManager, utteranceId, whichBatch

Constructor Summary

BatchNISTRecognizer()

BatchNISTRecognizer(Recognizer recognizer, java.util.List<DataProcessor> inputDataProcessors, java.lang.String ctlFile, java.lang.String dataDir, java.lang.String refFile, java.lang.String ctmFile, int bitsPerSample, int samplesPerSecond, int framesPerSecond, int channelCount)

Method Summary

void	decode()
private int	dumpBestPath(java.io.DataOutputStream out, BatchNISTRecognizer.CTLUtterance utt, Token token)
protected void	handleResult(java.io.DataOutputStream out, BatchNISTRecognizer.CTLUtterance utt, Result result)
static byte[]	hex2Binary(java.lang.String spelling)
private static int	hexToByte(char c)
static void	main(java.lang.String[] argv)
void	newProperties(PropertySheet ps) This method is called when this configurable component needs to be reconfigured.
protected void	setInputStream(BatchNISTRecognizer.CTLUtterance utt)

Methods inherited from class edu.cmu.sphinx.tools.batch.BatchModeRecognizer

addCommands, decode, recognize, setBatchFile, setInputStream, shell, start, stop

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

ctlFile

protected java.lang.String ctlFile

dataDir

protected java.lang.String dataDir

refFile

protected java.lang.String refFile

ctmFile

protected java.lang.String ctmFile

bitsPerSample

protected int bitsPerSample

samplesPerSecond

protected int samplesPerSecond

framesPerSecond

protected int framesPerSecond

channelCount

protected int channelCount

bytesPerFrame

protected int bytesPerFrame

PROP_DATA_DIR

@S4String(defaultValue="")
public static final java.lang.String PROP_DATA_DIR

The property that specifies the file containing the corpus utterance audio

See Also:

Constant Field Values

PROP_CTL_FILE

@S4String(defaultValue="")
public static final java.lang.String PROP_CTL_FILE

The property that specifies the file containing the corpus utterance audio

See Also:

Constant Field Values

PROP_REF_FILE

@S4String(defaultValue="")
public static final java.lang.String PROP_REF_FILE

The property that specifies the file containing the transcripts of the corpus

See Also:

Constant Field Values

PROP_CTM_FILE

@S4String(defaultValue="")
public static final java.lang.String PROP_CTM_FILE

The property that specifies the the directory where the output XXX files should be placed

See Also:

Constant Field Values

PROP_BITS_PER_SAMPLE

@S4Integer(defaultValue=16)
public static final java.lang.String PROP_BITS_PER_SAMPLE

The sphinx properties that specify the format of the PCM audio in the data file

See Also:

Constant Field Values

PROP_CHANNEL_COUNT

@S4Integer(defaultValue=1)
public static final java.lang.String PROP_CHANNEL_COUNT

See Also:

Constant Field Values

PROP_SAMPLES_PER_SECOND

@S4Integer(defaultValue=16000)
public static final java.lang.String PROP_SAMPLES_PER_SECOND

See Also:

Constant Field Values

PROP_FRAMES_PER_SECOND

@S4Integer(defaultValue=100)
public static final java.lang.String PROP_FRAMES_PER_SECOND

See Also:

Constant Field Values

Constructor Detail

BatchNISTRecognizer

public BatchNISTRecognizer(Recognizer recognizer,
                           java.util.List<DataProcessor> inputDataProcessors,
                           java.lang.String ctlFile,
                           java.lang.String dataDir,
                           java.lang.String refFile,
                           java.lang.String ctmFile,
                           int bitsPerSample,
                           int samplesPerSecond,
                           int framesPerSecond,
                           int channelCount)

BatchNISTRecognizer

public BatchNISTRecognizer()

Method Detail

newProperties

public void newProperties(PropertySheet ps)
                   throws PropertyException

Description copied from interface: Configurable

This method is called when this configurable component needs to be reconfigured.

Specified by:

newProperties in interface Configurable

Overrides:

newProperties in class BatchModeRecognizer

Parameters:

ps - a property sheet holding the new data

Throws:

PropertyException - if there is a problem with the properties.

setInputStream

protected void setInputStream(BatchNISTRecognizer.CTLUtterance utt)
                       throws java.io.IOException

Throws:

java.io.IOException

decode

public void decode()

handleResult

protected void handleResult(java.io.DataOutputStream out,
                            BatchNISTRecognizer.CTLUtterance utt,
                            Result result)
                     throws java.io.IOException

Throws:

java.io.IOException

dumpBestPath

private int dumpBestPath(java.io.DataOutputStream out,
                         BatchNISTRecognizer.CTLUtterance utt,
                         Token token)
                  throws java.io.IOException

Throws:

java.io.IOException

hex2Binary

public static byte[] hex2Binary(java.lang.String spelling)

hexToByte

private static int hexToByte(char c)

main

public static void main(java.lang.String[] argv)