edu.cmu.sphinx.linguist.acoustic.tiedstate.trainer

Class TrainerAcousticModel

java.lang.Object

edu.cmu.sphinx.linguist.acoustic.tiedstate.TiedStateAcousticModel

edu.cmu.sphinx.linguist.acoustic.tiedstate.trainer.TrainerAcousticModel

All Implemented Interfaces:

AcousticModel, Configurable

public class TrainerAcousticModel
extends TiedStateAcousticModel

Represents the generic interface to the Acoustic Model for sphinx4

Field Summary

Fields

Modifier and Type	Field and Description
static int	ALL_MODELS Flag indicating all models should be operated on.
static java.lang.String	PROP_FORMAT_SAVE The save format for the acoustic model data.
static java.lang.String	PROP_MC_FLOOR Mixture component score floor.
static java.lang.String	PROP_MW_FLOOR Mixture weight floor.
static java.lang.String	PROP_VARIANCE_FLOOR
java.lang.String	saveFormat
static java.lang.String	SAVER

Fields inherited from class edu.cmu.sphinx.linguist.acoustic.tiedstate.TiedStateAcousticModel

loader, logger, name, PROP_LOADER, PROP_UNIT_MANAGER, PROP_USE_COMPOSITES, unitManager

Fields inherited from interface edu.cmu.sphinx.linguist.acoustic.AcousticModel

PROP_LOCATION

Constructor Summary

Constructors

Constructor and Description
TrainerAcousticModel()
TrainerAcousticModel(Loader loader, UnitManager unitManager, boolean useComposites, Saver saver, java.lang.String saveFormat)

Method Summary

Modifier and Type	Method and Description
void	accumulate(int index, TrainerScore[] trainerScore) Accumulate the current TrainerScore into the buffers.
void	accumulate(int index, TrainerScore[] trainerScore, TrainerScore[] nextTrainerScore) Accumulate the current TrainerScore into the buffers.
void	load() Loads the acoustic models.
void	newProperties(PropertySheet ps) This method is called when this configurable component needs to be reconfigured.
float	normalize() Normalize the buffers and update the models.
void	resetBuffers() Resets the buffers.
void	save(java.lang.String name) Saves the acoustic model with a given name and format
void	updateLogLikelihood() Update the log likelihood.

Methods inherited from class edu.cmu.sphinx.linguist.acoustic.tiedstate.TiedStateAcousticModel

allocate, deallocate, getCompositeSenoneSequence, getContextIndependentUnitIterator, getHMMIterator, getLeftContextSize, getName, getProperties, getRightContextSize, getSenone, logInfo, lookupNearestHMM

Methods inherited from class java.lang.Object

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

Field Detail

SAVER

@S4Component(type=Saver.class)
public static final java.lang.String SAVER

See Also:

Constant Field Values

PROP_VARIANCE_FLOOR

@S4Double(defaultValue=9.999999747378752E-5)
public static final java.lang.String PROP_VARIANCE_FLOOR

See Also:

Constant Field Values

PROP_MC_FLOOR

@S4Double(defaultValue=0.0)
public static final java.lang.String PROP_MC_FLOOR

Mixture component score floor.

See Also:

Constant Field Values

PROP_MW_FLOOR

@S4Double(defaultValue=1.0000000116860974E-7)
public static final java.lang.String PROP_MW_FLOOR

Mixture weight floor.

See Also:

Constant Field Values

PROP_FORMAT_SAVE

@S4String(defaultValue="sphinx3.binary")
public static final java.lang.String PROP_FORMAT_SAVE

The save format for the acoustic model data. Current supported formats are:

See Also:

Constant Field Values

saveFormat

public java.lang.String saveFormat

ALL_MODELS

public static final int ALL_MODELS

Flag indicating all models should be operated on.

See Also:

Constant Field Values

Constructor Detail

TrainerAcousticModel

public TrainerAcousticModel(Loader loader,
                            UnitManager unitManager,
                            boolean useComposites,
                            Saver saver,
                            java.lang.String saveFormat)
                     throws java.io.IOException

Throws:

java.io.IOException

TrainerAcousticModel

public TrainerAcousticModel()

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 TiedStateAcousticModel

Parameters:

ps - a property sheet holding the new data

Throws:

PropertyException - if there is a problem with the properties.

save

public void save(java.lang.String name)
          throws java.io.IOException

Saves the acoustic model with a given name and format

Parameters:

name - the name of the acoustic model

Throws:

java.io.IOException - if the model could not be loaded

java.io.FileNotFoundException - if the model does not exist

load

public void load()
          throws java.io.IOException,
                 java.io.FileNotFoundException

Loads the acoustic models. This has to be explicitly requested in this class.

Throws:

java.io.IOException - if the model could not be loaded

java.io.FileNotFoundException - if the model does not exist

resetBuffers

public void resetBuffers()

Resets the buffers.

accumulate

public void accumulate(int index,
                       TrainerScore[] trainerScore,
                       TrainerScore[] nextTrainerScore)

Accumulate the current TrainerScore into the buffers.

Parameters:

index - the current index into the TrainerScore vector

trainerScore - the TrainerScore in the current frame

nextTrainerScore - the TrainerScore in the next frame

accumulate

public void accumulate(int index,
                       TrainerScore[] trainerScore)

Accumulate the current TrainerScore into the buffers.

Parameters:

index - the current index into the TrainerScore vector

trainerScore - the TrainerScore

updateLogLikelihood

public void updateLogLikelihood()

Update the log likelihood. This should be called at the end of each utterance.

normalize

public float normalize()

Normalize the buffers and update the models.

Returns:

the log likelihood for the whole training set