libak  0.4.0
Data Structures | Functions
MLLR adaptation
Algorithms

Data Structures

struct  akMLLRrc
 Used to assign states to regression classes. More...

Functions

void ak_mllr_free (akMLLR *mllr)
 Frees memory.
akCModelak_mllr_get_adapted_model (akMLLR *mllr, akCModel *cmodel)
 Adapts the model.
void ak_mllr_init (akMLLR *mllr)
 Initializes the MLLR transformation matrices.
akMLLR * ak_mllr_new (akCModel *cmodel, const akMLLRrc *rcs, const unsigned int length_rcs, char **err)
 Get an akMLLR.
void ak_mllr_update_matrices (akMLLR *mllr, const akCMlogCounts *counts, const akCModel *cmodel)
 Update the transformation matrices.

Function Documentation

void ak_mllr_free ( akMLLR *  mllr)

Frees memory.

Frees the memory allocated for the MLLR structure.

Parameters:
mllrThe MLLR structure.
akCModel* ak_mllr_get_adapted_model ( akMLLR *  mllr,
akCModel cmodel 
)

Adapts the model.

This function adapts the character models passed using the current MLLR transformation matrices. The character models must have the same structure than the original the character models which were used to make the MLLR structure.

Parameters:
mllrThe MLLR structure.
cmodelThe character models to adapt. If NULL, then original character models are copied and adapted.
Returns:
The adapted character models.
void ak_mllr_init ( akMLLR *  mllr)

Initializes the MLLR transformation matrices.

This function initializes the MLLR transformation matrices. The A matrix is initialized to the identity matrix, and the b vector is initialized with zeroes.

Parameters:
mllrThe MLLR structure.
akMLLR* ak_mllr_new ( akCModel cmodel,
const akMLLRrc rcs,
const unsigned int  length_rcs,
char **  err 
)

Get an akMLLR.

This function creates a MLLR structure used to adapt models and transform features using the MLLR technique. Only DGaussian based types are supported. This function calls 'ak_mllr_init'.

Parameters:
cmodelThe character models.
rcsAn array used to assign states to regression classes. By default all classes are assigned to the regression class 0. If NULL then all states are assigned to the regression class 0.
length_rcsThe size of rcs.
errPointer to string variable. If not NULL an error message is allocated in the variable in case of error.
Returns:
MLLR structure or NULL in case of error.
void ak_mllr_update_matrices ( akMLLR *  mllr,
const akCMlogCounts counts,
const akCModel cmodel 
)

Update the transformation matrices.

This function recalculates the MLLR transformation matrices using the provided logarithmic counts and the parameters of the character models used to estimate the counts. Counts must be created using 'ak_cmlogcounts_new_mllr', and the character models must have the same structure than the structure of the character models used to create the MLLR structure.

Parameters:
mllrThe MLLR structure.
countsLogarithmic counts related to the character models of the CMLLR structure. The counts must be created using 'ak_cmlogcounts_new_mllr'.
cmodelThe character models parameters used to estimate the counts.
 All Data Structures Variables