libTLK  1.3.1
Data Fields
tLFeatureExtractor Struct Reference

Feature extractor. More...

Data Fields

tLTypeFeatures ftype
 Type of the features.
int outfreq
 Output frequency.
double wduration
 Duration for each extraction window in seconds.
int numchans
 Number of channels used for the Mel filter bank.
int lofreq
 Low frequency cut-off for filterbank analysis.
int hifreq
 High frequency cut-off for filterbank analysis.
int numceps
 Number of cepstral features (only for MFCC).
tLBool zeromean
 Removes the DC mean from the source.
double preemcoef
 Coefficient used to pre-emphasize the signal ([0,1[).
tLBool usehamming
 Apply Hamming windowing.
tLBool usepower
 Uses the power in filterbank analysis.
int ceplifter
 Value used to lifter the cepstra.
tLBool energy
 Calculates the energy and normalizes it.
int deltawindow
 Width of the window used to calculate delta coefficients.
tLBool cep0
 Add Cepstral 0.
tLBool meannorm
 Normalize the vectors substracting the mean.
int accwindow
 Width of the window used to calculate acceleration coefficients.

Detailed Description

Feature extractor.

This structure is used for feature extraction. Public fields could be modified at any time.


Field Documentation

Width of the window used to calculate acceleration coefficients.

A 0 value disables the calculation of acceleration coeeficients. Default value 2.

Add Cepstral 0.

Default value TRUE.

Value used to lifter the cepstra.

Default 22.

Width of the window used to calculate delta coefficients.

A 0 value disables the calculation of delta coefficients. Default value 2.

Calculates the energy and normalizes it.

Default TRUE.

tLTypeFeatures tLFeatureExtractor::ftype

Type of the features.

Default TL_MFCC.

High frequency cut-off for filterbank analysis.

Default INT_MAX.

Low frequency cut-off for filterbank analysis.

Default 0.

Normalize the vectors substracting the mean.

Default value TRUE.

Number of cepstral features (only for MFCC).

Default 12.

Number of channels used for the Mel filter bank.

Default 26.

Output frequency.

Default 100Hz.

Coefficient used to pre-emphasize the signal ([0,1[).

Default 0.97.

Apply Hamming windowing.

Default TRUE.

Uses the power in filterbank analysis.

Default TRUE.

Duration for each extraction window in seconds.

Default 0.025 (25ms).

Removes the DC mean from the source.

Default TRUE.

 All Data Structures Variables