SEGRec Version 1.0: SEGLib/segV3Compat.h File Reference

Go to the source code of this file.


Functions
int	MakeStrictLRTopology (SEG_MODEL_SET *modelSet)
	Force models to maintain strict left-to-right topology.
int	MLCheckModelSet (SEG_MODEL_SET *modelSet)
	Check the consistency of model parameter set.
int	MLReadModelSet (char modelList, SEG_MODEL_SET modelSet, int *longestDur)
	Read a set of multi-level segmental-HMMs in the file of model name list.
int	MLWriteModelSet (SEG_MODEL_SET modelSet, char dirName)
	Write a set of models to a file.
double	ViterbiDecoder (double spData, int sT, int eT, SEG_SYNTAX synt, SEG_MODEL_SET hmm, int maxDur, SEG_DP_ITEM dpMat, double segProb, int *breakPt)
	Generic Viterbi-type decorder for segmental HMMs.
int	FreeDPMat (SEG_DP_ITEM **dpMat, int numStates)
	Free memory space storing all the paths in the trellis.
int *	RecoverStateSeq (SEG_SYNTAX syn, SEG_DP_ITEM *dpHist, int numVecs, int numLabs)
	Recover the optimal state sequence from the path record of node.
SEG_STATE **	RecoverStateIds (SEG_SYNTAX syn, int *stateSeq, int numVecs)
	Recover the sequence of state identities from a state sequence.
SEG_ANN_DATA *	RecoverRecogSeq (SEG_SYNTAX syn, SEG_DP_ITEM **dpMat, int numVecs, int numLabs)
	Recover the model labels and their start and end times.
int	FreeStateSeq (int *stateSeq)
	Free memory space of state sequence.
int	WriteMLFData (SEG_ANN_DATA recOp, int nLabs, int winSize, FILE mlfFile, char labFileName)
	Generate and write new annotations from the classification/recognition output.
int	GetDurationLimits (SEG_MODEL_SET models, SEG_SYNTAX synt, int eStates, int minUttDur, int maxUttDur)
	Find the minimum and maximum possible durations for a given syntax.
void	segSetDecoderParams (SEG_DEC_PARAM dec_param)
	Set decoding parameters.
void	segGetDecoderParams (SEG_DEC_PARAM *dec_param)
	Get decoding parameters.
int	MLIncrementLinearHMMAccumulators (SEG_STATE *stateId, int stateSeq, SEG_SYNTAX syn, double **spData, int sT, int eT, int maxDur)
	Increment the reestimation bins, according to the current utterance alignment.
int	MLInitialiseHMMAccumulators (SEG_MODEL_SET *hmmSet)
	Initialise reestimation accumulators for the entire set of models.
int	MLReestSegModels (SEG_MODEL_SET *hmmSet, double minVar, double slopeRate, double varRate, double durRegParam)
	Reestimate linear segmental-HMM parameters.
int	GetSegVitConfig (char cFile, char hmmListFile, char trnFile, int maxIt, double stopDiff, char logFile, char mlfFile, int embedTrain, int maxDur, double minVar, char hmmOpDir, int runType, double slopeRate, double varRate, int verbose, char aMatFile, char iMatFile, char classFile, int opMats, char grammarFile, int LMScale, char *dumpPtr)
	Read SEGRec configuration.
int	ReadGrammar (char , SEG_MODEL_SET, double , double , double )
	Read bigram grammar (bigram posterior probability table).
int	MakeTrainingSyntax (SEG_UTT_LIST_ITEM, SEG_MODEL_SET, SEG_SYNTAX , int )
	Create 'forced alignment' syntax (for supervised training and classification).
int	MakeRecognitionSyntax (SEG_MODEL_SET, SEG_SYNTAX , int, double , double , double *, double)
	Create recognition syntax.
int	CheckSyn (SEG_MODEL_SET, SEG_SYNTAX *)
	Display the contents of syntax for debugging.
int	FreeSyntaxMemory (SEG_SYNTAX *, int)
	Free memory allocated to training/recognition syntax.

Detailed Description

Function Documentation

int CheckSyn	(	SEG_MODEL_SET	models,
		SEG_SYNTAX *	synt
	)

Display the contents of syntax for debugging.

Parameters:

`[in]`	models	Model set (pointer to a variable of struct modelSet)
`[in]`	synt	Syntax (pointer to a variable of struct syntax)

Returns:: Number of nodes in the syntax

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segCheckSyn() internally.

See also:: segCheckSyn()

int FreeDPMat	(	SEG_DP_ITEM **	dpMat,
		int	numStates
	)

Free memory space storing all the paths in the trellis.

Parameters:

`[in]`	dpMat	Paths in the trellis for back-tracing
`[in]`	numStates	Number of states

Return values:

NO_ERROR

Successful execution

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segFreeDecoder() internally.

See also:: segFreeDecoder()

int FreeStateSeq ( int * stateSeq )

Free memory space of state sequence.

Parameters:

[in] stateSeq State (index) sequence

Return values:

NO_ERROR

Successful execution

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segFreeStateSeq() internally.

See also:: segFreeStateSeq()

int FreeSyntaxMemory	(	SEG_SYNTAX *	synt,
		int	totStates
	)

Free memory allocated to training/recognition syntax.

Parameters:

`[in]`	synt	Syntax
`[in]`	totStates	Number of states in the syntax

Return values:

NO_ERROR

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segFreeSyntaxMemory() internally.

See also:: segFreeSyntaxMemory()

int GetDurationLimits	(	SEG_MODEL_SET	models,
		SEG_SYNTAX	synt,
		int	eStates,
		int *	minUttDur,
		int *	maxUttDur
	)

Find the minimum and maximum possible durations for a given syntax.

Parameters:

`[in]`	models	Model set
`[in]`	synt	Syntax
`[in]`	eStates	Number of emitting states in the syntax
`[in]`	minUttDur	Minimum utterance duration
`[in]`	maxUttDur	Maximum utterance duration

Return values:

	RTN_ERROR	Inconsistency in the models
	NO_ERROR	Successfully executed

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segGetDurationLimits() internally.

See also:: segGetDurationLimits()

int GetSegVitConfig	(	char *	cFile,
		char *	hmmListFile,
		char *	trnFile,
		int *	maxIt,
		double *	stopDiff,
		char *	logFile,
		char *	mlfFile,
		int *	embedTrain,
		int *	maxDur,
		double *	minVar,
		char *	hmmOpDir,
		int *	runType,
		double *	slopeRate,
		double *	varRate,
		int *	verbose,
		char *	aMatFile,
		char *	iMatFile,
		char *	classFile,
		int *	opMats,
		char *	grammarFile,
		int *	LMScale,
		char **	dumpPtr
	)

Read SEGRec configuration.

Parameters:

`[in]`	cFile	Name of configuration file
`[in,out]`	hmmListFile	Name of HMM list file
`[in,out]`	trnFile	Name of speech list file
`[in,out]`	maxIt	Maximum number of iterations for training
`[in,out]`	stopDiff	Fractional threshold to stop iterations
`[in,out]`	logFile	Name of log file
`[in,out]`	mlfFile	Name of MLF file
`[in,out]`	embedTrain	Flag switching supervised or embedded decoding
`[in,out]`	maxDur	Maximal segment-duration
`[in,out]`	minVar	Variance floor
`[in,out]`	hmmOpDir	Name of directory where models are stored
`[in,out]`	runType	Flag switching recognition or training
`[in,out]`	slopeRate	Update rate for slope parameter
`[in,out]`	varRate	Update rate for variance parameter
`[in,out]`	verbose	Level of reporting displayed during execution
`[in,out]`	aMatFile	Name of file storing acoustic features
`[in,out]`	iMatFile	Name of file storing articulatory features
`[in,out]`	classFile	Name of class file
`[in,out]`	opMats	Variable specifying additional debug info to be output to certain files
`[in,out]`	grammarFile	Name of grammar file
`[in,out]`	LMScale	Weight for prob from language model against acoustic probabilities
`[in,out]`	dumpPtr	Name of dump file

Return values:

	ERROR_FOPEN	File open error
	ERROR_ALLOC	Memory allocation error
	RTN_ERROR	Other errors
	otherwise	Number of parameters read

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segReadConfig() and segFreeConfig() internally.

See also:: segReadConfig() and segFreeConfig()

int MakeRecognitionSyntax	(	SEG_MODEL_SET	models,
		SEG_SYNTAX *	synt,
		int	recogType,
		double *	initProb,
		double *	exitProb,
		double **	biProb,
		double	lmff
	)

Create recognition syntax.

Parameters:

`[in]`	models	Set of models with model parameters
`[in,out]`	synt	Syntax to be built
`[in]`	recogType	Type of testing (RECOGNITION or CLASSIFICATION)
`[in]`	initProb	Entry costs for the models
`[in]`	exitProb	Exit costs for the models
`[in]`	biProb	N-gram posterior probabilities table
`[in]`	lmff	Scale factor for the language model

Return values:

	RTN_ERROR	Memory allocation error / Inconsistency between models
	otherwise	Number of nodes in the syntax

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segMakeRecognitionSyntax() internally.

Attention:: Use segMakeRecognitionSyntax() instead. The SEGVit compatible function MakeRecognitionSyntax() is no longer supplied, because the old interface cannot deal with the context-sensitive models of SEGRec.

See also:: segMakeRecognitionSyntax()

int MakeStrictLRTopology ( SEG_MODEL_SET * modelSet )

Force models to maintain strict left-to-right topology.

Parameters:

[in,out] modelSet Pointer to a struct variable for the model set

Return values:

NO_ERROR

Successful execution

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segMakeStrictLRTopology() internally.

See also:: segMakeStrictLRTopology()

int MakeTrainingSyntax	(	SEG_UTT_LIST_ITEM	utt,
		SEG_MODEL_SET	models,
		SEG_SYNTAX *	synt,
		int *	eStates
	)

Create 'forced alignment' syntax (for supervised training and classification).

Parameters:

`[in]`	utt	Utterance
`[in]`	models	Set of models with model parameters
`[in,out]`	synt	Syntax to be built
`[in,out]`	eStates	Number of emitting states

Return values:

	RTN_ERROR	Memory allocation error / Inconsistency between models
	otherwise	Number of nodes in the syntax

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segMakeTrainingSyntax() internally.

See also:: segMakeTrainingSyntax()

int MLCheckModelSet ( SEG_MODEL_SET * modelSet )

Check the consistency of model parameter set.

Parameters:

[in] modelSet Pointer to a struct variable for the model set

Return values:

	NO_ERROR	Successful execution
	RTN_ERROR	Inconsistency found in the set

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segCheckModelSet() internally.

See also:: segCheckModelSet()

int MLIncrementLinearHMMAccumulators	(	SEG_STATE **	stateId,
		int *	stateSeq,
		SEG_SYNTAX	syn,
		double **	spData,
		int	sT,
		int	eT,
		int	maxDur
	)

Increment the reestimation bins, according to the current utterance alignment.

Parameters:

`[in,out]`	stateId	Array of state IDs of the best state-sequence
`[in]`	stateSeq	The best state sequence
`[in]`	syn	Syntax
`[in]`	spData	Acoustic feature vectors of speech data
`[in]`	sT	Start time of the utterance
`[in]`	eT	End time of the utterance
`[in]`	maxDur	Maximum segment duration

Return values:

	RTN_ERROR	Memory allocation error
	NO_ERROR	Successfully executed

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segIncrementLinearHMMAccumulators() internally.

See also:: segIncrementLinearHMMAccumulators()

int MLInitialiseHMMAccumulators ( SEG_MODEL_SET * hmmSet )

Initialise reestimation accumulators for the entire set of models.

Parameters:

[in,out] hmmSet Model set

Returns:: NO_ERROR Successfully executed

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segInitialiseHMMAccumulators() internally.

See also:: segInitialiseHMMAccumulators()

int MLReadModelSet	(	char *	modelList,
		SEG_MODEL_SET *	modelSet,
		int *	longestDur
	)

Read a set of multi-level segmental-HMMs in the file of model name list.

Parameters:

`[in]`	modelList	Name of model-name list file
`[in,out]`	modelSet	Model set
`[in]`	longestDur	Maximal segment-duration

Return values:

	RTN_ERROR	Error
	otherwise	Dimension of the intermediate features

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segReadModelSet() internally.

: Logical and physical models are set to have the same labels and model parameters, because SEGVit cannot handle "tied list" for context-sensitive phones. Use segReadModelSet() instead, to deal with logical models different from physical models.

See also:: segReadModelSet()

int MLReestSegModels	(	SEG_MODEL_SET *	hmmSet,
		double	minVar,
		double	slopeRate,
		double	varRate,
		double	durRegParam
	)

Reestimate linear segmental-HMM parameters.

Parameters:

`[in,out]`	hmmSet	Model set
`[in]`	minVar	Variance floor
`[in]`	slopeRate	Update rate of slope in comparison to the mean
`[in]`	varRate	Update rate of variance in comparison to the mean
`[in]`	durRegParam	Regularisation parameter for the durational model

Return values:

	NO_ERROR	Successfully executed
	RTN_ERROR	Too small number of samples for the model

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segReestSegModels() internally.

See also:: segReestSegModels()

int MLWriteModelSet	(	SEG_MODEL_SET *	modelSet,
		char *	dirName
	)

Write a set of models to a file.

Parameters:

`[in]`	modelSet	Pointer to the model-set structure variable
`[in]`	dirName	Name of directory where the models are saved

Return values:

	RTN_ERROR	Memory allocation error / File write error
	NO_ERROR	Successful execution

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segWriteModelSet() internally. It outputs the current models to the specified directory.

See also:: segWriteModelSet()

int ReadGrammar	(	char *	gramFName,
		SEG_MODEL_SET	models,
		double **	initProb,
		double ***	biProb,
		double **	exitProb
	)

Read bigram grammar (bigram posterior probability table).

Parameters:

`[in]`	gramFName	Name of the grammar file
`[in]`	models	Model set
`[in,out]`	initProb	Entry costs for the models
`[in,out]`	biProb	N-gram posterior probabilities table
`[in,out]`	exitProb	Exit costs for the models

Return values:

	RTN_ERROR	File open/read error / Memory allocation error
	NO_ERROR	Successfully executed

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segReadGrammar() internally.

See also:: segReadGrammar()

SEG_ANN_DATA* RecoverRecogSeq	(	SEG_SYNTAX	syn,
		SEG_DP_ITEM **	dpMat,
		int	numVecs,
		int	numLabs
	)

Recover the model labels and their start and end times.

Parameters:

`[in]`	syn	Syntax
`[in]`	dpMat	Paths in the trellis for back-tracing
`[in]`	numVecs	Number of frames in the sequence
`[in]`	numLabs	Number of models in the sequence

Return values:

	NULL	Memory allocation error
	otherwise	Pointer to the array of struct segAnnData variables each of which contains a label, and start and end times.

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segRecoverRecogSeq() internally.

See also:: segRecoverRecogSeq()

SEG_STATE** RecoverStateIds	(	SEG_SYNTAX	syn,
		int *	stateSeq,
		int	numVecs
	)

Recover the sequence of state identities from a state sequence.

Parameters:

`[in]`	syn	Syntax
`[in]`	stateSeq	State sequence
`[in]`	numVecs	Number of frames in the utterance

Return values:

	NULL	Memory allocation error
	otherwise	Pointer to the array of struct segStateInfo pointers

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segRecoverStateIds() internally.

See also:: segRecoverStateIds()

int* RecoverStateSeq	(	SEG_SYNTAX	syn,
		SEG_DP_ITEM **	dpHist,
		int	numVecs,
		int *	numLabs
	)

Recover the optimal state sequence from the path record of node.

Parameters:

`[in]`	syn	Syntax
`[in]`	dpHist	Paths in the trellis for back-tracing
`[in]`	numVecs	Number of frames in the utterance
`[in,out]`	numLabs	Number of models in the sequence recovered

Return values:

	NULL	Memory allocation error
	Otherwise	Pointer to the array of state sequence recovered

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segRecoverStateSeq() internally.

See also:: segRecoverStateSeq()

void segGetDecoderParams ( SEG_DEC_PARAM * dec_param )

Get decoding parameters.

Parameters:

[in] dec_param Struct variable of decoding parameters

Returns:: No return value

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function provides the way of setting decoding parameters in the framework of SEGVit ver.3.4, where those parameters cannot be given as an input argument of the decoder.

void segSetDecoderParams ( SEG_DEC_PARAM dec_param )

Set decoding parameters.

Parameters:

[in] dec_param Struct variable of decoding parameters

Returns:: No return value

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function provides the way of setting decoding parameters in the framework of SEGVit ver.3.4, where those parameters cannot be given as an input argument of the decoder.

double ViterbiDecoder	(	double **	spData,
		int	sT,
		int	eT,
		SEG_SYNTAX	synt,
		SEG_MODEL_SET	hmm,
		int	maxDur,
		SEG_DP_ITEM ***	dpMat,
		double *	segProb,
		int *	breakPt
	)

Generic Viterbi-type decorder for segmental HMMs.

Parameters:

`[in]`	spData	Acoustic feature vectors
`[in]`	sT	Utterance start time (frame No)
`[in]`	eT	Utterance end time (frame No)
`[in]`	synt	Syntax
`[in]`	hmm	Model set
`[in]`	maxDur	Maximum segment-duration
`[out]`	dpMat	Paths in the trellis for back-tracing
`[out]`	segProb	Output probabilities for SHMMs
`[in]`	breakPt	Break point attributes

Return values:

	RTN_ERROR	Memory allocation error
	Otherwise	The best cost (min probability) computed

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segViterbiDecoder() internally.

See also:: segViterbiDecoder

int WriteMLFData	(	SEG_ANN_DATA	recOp,
		int	nLabs,
		int	winSize,
		FILE *	mlfFile,
		char *	labFileName
	)

Generate and write new annotations from the classification/recognition output.

Parameters:

`[in]`	recOp	Array of struct annData variables
`[in]`	nLabs	Number of labels in the utterance
`[in]`	winSize	Size of the window (ms) used in processing speech data
`[in]`	mlfFile	File pointer to a MLF file
`[in]`	labFileName	Name of the label file

Return values:

	RTN_ERROR	File write error
	NO_ERROR	Successful execution

Author:: Yoshinori Shiga @ CVSSP, University of Surrey

Note:: This function is a wrapper compatible with SEGVit ver.3.4, and calls SEGRec's segWriteMLFData() internally.

See also:: segWriteMLFData()

segV3Compat.h File Reference

Functions

Detailed Description

Function Documentation