Posterior probability of labels (phonemes) estimator based on artificial neural networks. More...
#include <bsapi.h>
Inheritance diagram for SNNetPosteriorEstimatorI:
Public Member Functions | |
| virtual bool BSAPI_METHOD | SetSystem (const char *sys="1BT_DCT")=0 |
| virtual void BSAPI_METHOD | SetSofteningFunc (psoft_func_type pfunc, float v1, float v2, float v3)=0 |
| virtual bool BSAPI_METHOD | SetSofteningFuncStr (const char *pSofteningFunc="none 0 0 0")=0 |
| virtual void BSAPI_METHOD | SetTrapLen (int trapLen=31)=0 |
| virtual void BSAPI_METHOD | SetUseHamming (bool useHamming=false)=0 |
| virtual void BSAPI_METHOD | SetNBanks (int nBanks=15)=0 |
| virtual void BSAPI_METHOD | SetAddC0 (bool addC0=true)=0 |
| virtual void BSAPI_METHOD | SetInvertBeforeMerger (bool invertBeforeMerger=true)=0 |
| virtual void BSAPI_METHOD | SetFrameStep (int frameStep=1)=0 |
| virtual void BSAPI_METHOD | SetBatchSize (int batchSize=1)=0 |
| virtual void BSAPI_METHOD | SetOutBatchSize (int outBatchSize=0)=0 |
| virtual void BSAPI_METHOD | SetReadTransforms (bool readTransforms=true)=0 |
| Enable reading of input feature transforms during initialization of the module by Load. | |
| virtual void BSAPI_METHOD | SetUseFastExp (bool useFastExp=false)=0 |
| virtual bool BSAPI_METHOD | Load (const char *pDir)=0 |
| Load the neural networks and initialize the system. | |
Detailed Description
Posterior probability of labels (phonemes) estimator based on artificial neural networks.
Member Function Documentation
| virtual bool BSAPI_METHOD SNNetPosteriorEstimatorI::Load | ( | const char * | pDir | ) | [pure virtual] |
Load the neural networks and initialize the system.
- Parameters:
-
pDir directory of data files: $dir/weights/band[i].weights - band weights in the ICSI QuickNet format (ascii) $dir/weights/merger.weights - merger weights in the ICSI QuickNet format (ascii) $dir/norms/band[i].norms - band mean and variance normalization coefficients in the ICSI QuickNEr format (ascii) $dir/norms/merger.norms - band mean and variance normalization coefficients in the ICSI QuickNEr format (ascii) $dir/weights/band[i].nbin - band weights and norms in a binary format. The file is generated during the first run of initialization. The *.weights and *.norms files are not necessary any more. $dir/weights/merger.nbin - band weights and norms in a binary format $dir/windows/band[i].window - temporal pattern weighting windows in the ascii format $dir/windows/band[i].wbin - temporal pattern weighting windows in the binary (HTK) format $dir/transforms/band[i].trans - linear transformations in the ascii format $dir/transforms/bands[i].tbin - linear thransformations in the binary (HTK) format The band is an alias for one frequency band (1BT, 3BT - the trap systems) or for one context (LCRC system)
- Returns:
- true it the module was initialized correctly
| virtual void BSAPI_METHOD SNNetPosteriorEstimatorI::SetReadTransforms | ( | bool | readTransforms = true | ) | [pure virtual] |
Enable reading of input feature transforms during initialization of the module by Load.
- Parameters:
-
readTransforms true to enable reading of transform
The documentation for this class was generated from the following file:
