|
|
#include <vcd.h>
Data Fields | |
| int_least16_t | unnormMixScore |
| int_least16_t | unnormFillerScore |
| int_least16_t | unnormStayScore |
| int_least16_t | unnormProgScore |
| int_least16_t | unnormSkipScore |
| int_least16_t | iMaxDelta |
| int_least16_t | hmmTransWgt |
| int_least16_t | oovMult |
| int_least16_t | scrThreshMult |
| int_least16_t | minModelStates |
| int_least16_t | maxModelStates |
| char const * | keyword |
| char const * | filler |
| bool | waitCommit |
| int_least16_t | sampleRate |
| int_least16_t | frameLength |
| int_least16_t | vadAlpha |
| int_least16_t | vadSigSNRThresh |
A structure holding run-time parameters to tune the recognizer. These parameters are used to alter operation and tune recognizer performance. These parameters are initialized to default values when the VCD recognizer is initialized, and can be retrieved and set using the VCD_getParams and VCD_setParams function calls.
| int_least16_t VCD_ParamsStructDef::unnormMixScore |
Mixture loop transition ln probability in Q15 ( <= 0 )
| int_least16_t VCD_ParamsStructDef::unnormFillerScore |
Filler mixture model loop transition ln probability in Q15 ( <= 0 )
| int_least16_t VCD_ParamsStructDef::unnormStayScore |
Left/Right model loop transition ln probability in Q15 ( <= 0 )
| int_least16_t VCD_ParamsStructDef::unnormProgScore |
Left/Right model progress transition ln probability in Q15 ( <= 0 )
| int_least16_t VCD_ParamsStructDef::unnormSkipScore |
Left/Right model skip transition ln probability in Q15 ( <= 0 )
| int_least16_t VCD_ParamsStructDef::iMaxDelta |
Maximum 1/delta score to cause rejection in Q11
0 < iMaxDelta < 16
| int_least16_t VCD_ParamsStructDef::hmmTransWgt |
Relative weighting between acoustic and transition scores in Q11
0.0625 < HMM_TRANS_WGT < 16
| int_least16_t VCD_ParamsStructDef::oovMult |
Out of vocabulary score difference multiplier in Q12
0 < oovMult < 8
| int_least16_t VCD_ParamsStructDef::scrThreshMult |
Out of vocabulary score threshold multiplier in Q12
0 < oovMult < 8
| int_least16_t VCD_ParamsStructDef::minModelStates |
Minimum model states used to ensure quality
20 <= minModelStates <= maxModelStates
| int_least16_t VCD_ParamsStructDef::maxModelStates |
Maximum model states to limit size Must be greater than minModelStates. Maximum depends on platform.
| char const* VCD_ParamsStructDef::keyword |
Pointer to keyword model name If it exists, early commit the one model only. Initially not set to a valid model
| char const* VCD_ParamsStructDef::filler |
Mixture type model to use as filler model Initially set to use the unique provided Filler model. If model exists use it as a filler model
| bool VCD_ParamsStructDef::waitCommit |
Flag indicating when to commit recongition result. If set, then VCD will wait until speech is over before committing a recognition result.
| int_least16_t VCD_ParamsStructDef::sampleRate |
Sample rate required in Hz; only gettable, not settable
| int_least16_t VCD_ParamsStructDef::frameLength |
Frame length required in samples; only gettable, not settable
| int_least16_t VCD_ParamsStructDef::vadAlpha |
VAD SNR duration response parameter. This parameter adjusts sensitivity to speech input duration above background before the search engine will start.
| int_least16_t VCD_ParamsStructDef::vadSigSNRThresh |
VAD SNR threshold. This parameter adjusts sensitivity to speech input level above background before the search engine will start.
