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

Class TrainerScore

java.lang.Object

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

public class TrainerScore
extends java.lang.Object

Used to transfer data from the trainer to the acoustic model

Constructor Summary

Constructors

Constructor and Description
TrainerScore(Data feature, float probability, HMMState state) Creates a new buffer
TrainerScore(Data feature, float probability, HMMState state, float[] logComponentProb) Creates a new buffer
TrainerScore(Data feature, float probability, HMMState state, float logAlpha, float[] logComponentProb) Creates a new buffer
TrainerScore(Data feature, float probability, HMMState state, float logAlpha, float logBeta, float[] logComponentProb) Creates a new buffer
TrainerScore(Data feature, float probability, int senone) Creates a new TrainerScore

Method Summary

Modifier and Type	Method and Description
float	getAlpha() Retrieves the forward probability.
float	getBeta() Retrieves the backward probability.
float[]	getComponentGamma() Retrieves the mixture component a posteriori probability.
Data	getData() Retrieves the Data.
float	getGamma() Retrieves the a posteriori probability.
static float	getLogLikelihood() Retrieves the utterance's log likelihood
float	getScalingFactor() Retrieves the scaling factor.
float	getScore() Retrieves the probability.
Senone	getSenone() Retrieves the senone.
int	getSenoneID() Retrieves the senone ID.
HMMState	getState() Retrieves the HMM state.
void	setAlpha(float logAlpha) Sets the forward probability.
void	setBeta(float logBeta) Sets the backward probability.
void	setGamma() Computes the a posteriori probability.
static void	setLogLikelihood(float likelihood) Sets the overall likelihood.
void	setScalingFactor(float logScalingFactor) Sets the scaling factor for this frame..

Methods inherited from class java.lang.Object

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

Constructor Detail

TrainerScore

public TrainerScore(Data feature,
                    float probability,
                    int senone)

Creates a new TrainerScore

Parameters:

feature - the current feature

probability - the score for the current frame

senone - the id for the current senone

TrainerScore

public TrainerScore(Data feature,
                    float probability,
                    HMMState state,
                    float logAlpha,
                    float logBeta,
                    float[] logComponentProb)

Creates a new buffer

Parameters:

feature - the current feature

probability - the score for the current frame

state - the HMMState for this score object

logAlpha - the forward probability

logBeta - the backward probability

logComponentProb - the mixture component a posteriori probabilities

TrainerScore

public TrainerScore(Data feature,
                    float probability,
                    HMMState state,
                    float logAlpha,
                    float[] logComponentProb)

Creates a new buffer

Parameters:

feature - the current feature

probability - the score for the current frame

state - the HMMState for this score object

logAlpha - the forward probability

logComponentProb - the mixture component a posteriori probabilities

TrainerScore

public TrainerScore(Data feature,
                    float probability,
                    HMMState state,
                    float[] logComponentProb)

Creates a new buffer

Parameters:

feature - the current feature

probability - the score for the current frame

state - the HMMState for this score object

logComponentProb - the mixture component a posteriori probabilities

TrainerScore

public TrainerScore(Data feature,
                    float probability,
                    HMMState state)

Creates a new buffer

Parameters:

feature - the current feature

probability - the score for the current frame

state - the HMMState for this score object

Method Detail

getData

public Data getData()

Retrieves the Data.

Returns:

the Data

getScore

public float getScore()

Retrieves the probability.

Returns:

the probability

getAlpha

public float getAlpha()

Retrieves the forward probability.

Returns:

the forward log probability

getBeta

public float getBeta()

Retrieves the backward probability.

Returns:

the backward log probability

getLogLikelihood

public static float getLogLikelihood()

Retrieves the utterance's log likelihood

Returns:

the log likelihood

getGamma

public float getGamma()

Retrieves the a posteriori probability.

Returns:

the a posteriori log probability

getComponentGamma

public float[] getComponentGamma()

Retrieves the mixture component a posteriori probability.

Returns:

the a posteriori log probabilities

getScalingFactor

public float getScalingFactor()

Retrieves the scaling factor.

Returns:

the scaling factor

setLogLikelihood

public static void setLogLikelihood(float likelihood)

Sets the overall likelihood.

Parameters:

likelihood - the log likelihood of the whole utterance

setAlpha

public void setAlpha(float logAlpha)

Sets the forward probability.

Parameters:

logAlpha - the forward log probability

setBeta

public void setBeta(float logBeta)

Sets the backward probability.

Parameters:

logBeta - the backward log probability

setGamma

public void setGamma()

Computes the a posteriori probability. This is the product of the current alpha and beta, or the summation of the current logAlpha and logBeta. The current beta is updated in the object, and the current alpha and beta are used.

setScalingFactor

public void setScalingFactor(float logScalingFactor)

Sets the scaling factor for this frame..

Parameters:

logScalingFactor - the log scaling factor

getSenoneID

public int getSenoneID()

Retrieves the senone ID.

Returns:

the senone ID

getSenone

public Senone getSenone()

Retrieves the senone.

Returns:

the senone

getState

public HMMState getState()

Retrieves the HMM state.

Returns:

the HMM state