|
SPRAAK
|
All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Groups Pages
|
SPRAAK
|
feature space MLLR More...
feature space MLLR
spr_fv_gauss [-DIAG](flag: diagonal elements only) [-SEG](flag: segmentation mode) [-cov output covariance matrices] [-stats collected statistics] [-a matrix A] [-b vector B] [-go new set of gaussians] [-add_cov incremental update of full cov gauss] [-add_stats incremental update of the compact fMLLR statistics] [-vlen vlen new preproc](0) [-iter max fMLLR iterations](-1) [-map MAP update](-1.0) [-beam beam search info] <-c Corpus> [-range b:e](0:-1) [-ssp script](SPR_BSS_DEV_NULL) [-obs ObsDir] [-suffix FileSuffix](sam) [-arcd FileName] <-h FileName> [-g FileName] [-sel FileName] [-am_opt Options] [-top_n Number(s)](0) [-rmg rmg_params](no) [-LMout Value](-100) [-NOGS](flag: no gauss sel) <-ci unit description files> [-cd unit description files] <-d dictionary> [-unwind unwind format] (-add==-add_cov) (-u==-ci)
| -DIAGflag | diagonal elements only Only calculate the diagonal elements of the covariance matrices |
| -SEGflag | segmentation mode Work in segmentation mode, i.e. read the segmentation from the corpus/segmentation file instead of doing Viterbi alignment |
| -cov<em>output | covariance matrices File to write the newly calculated (full) covariance matrices (gaussians) to. |
| -stats<em>collected | statistics File to write the compact fMLLR statistics (K and G matrices) to. |
| -a<em>matrix | A File to write the transformation matrix A (Y=AX+B) to. |
| -b<em>vector | B File to write the transformation vector B (Y=AX+B) to. |
| -go<em>new | set of gaussians File to write the new diagonal gaussian to. The new gaussian set was obtained using both fMLLR and MAP updates (requires the full covariance gaussians as input, not the compact fMLLR statistics; see option '-cov'). |
| -add_cov<em>incremental | update of full cov gauss Pre-load the given covariance matrices (incremental updates). Multiples file can be specified separated by ' .AND. ' (merging of results). |
| -add_stats<em>incremental | update of the compact fMLLR statistics Pre-load the given compact fMLLR statistics (incremental updates). Multiples file can be specified separated by ' .AND. ' (merging of results). |
| -vlen<em>vlen | new preproc Use a new preprocessing (different than the one used to train the model) to estimate the full covariance gaussians and/or fMLLR+HLDA transformation. The first elements in the output vector returned by the preprocessing script constitute the normal observation vector (used for the aligment and for the distrubution of the vector over the different gaussians in the mixture). The <vlen> last elements are the new preprocessing (e.g. input for the fMLLR+HLDA transformation). |
| -iter<em>max | fMLLR iterations Upper limit on the number of iterations when calculating the fMLLR transformation. |
| -map<em>MAP | update Combine fMLLR and MAP updates, the prior weight of the fMLLR gaussians is <map> frames. |
| -beam<em>beam | search info See load_search_space() for more details |
| -c<em>Corpus</em><a | name="spr_fv_gauss.c" class="el"> File with corpus entries or segmentations. |
| -range<em>b:e</em><a | name="spr_fv_gauss.range" class="el"> Optional begin and end entry the corpus/segmentation file. Counting starts at 0. |
| -ssp<em>script</em><a | name="spr_fv_gauss.ssp" class="el"> The signal processing script used to preprocess the input data. |
| -obs<em>ObsDir</em><a | name="spr_fv_gauss.obs" class="el"> Observation directory name. |
| -suffix<em>FileSuffix</em><a | name="spr_fv_gauss.suffix" class="el"> File suffix of the observation files (without leading '.'). |
| -arcd<em>FileName</em><a | name="spr_fv_gauss.arcd" class="el"> Unit file name (.arcd or .cd format). |
| -h<em>FileName</em><a | name="spr_fv_gauss.h" class="el"> The input HMM file. |
| -g<em>FileName</em><a | name="spr_fv_gauss.g" class="el"> The input MVG file (gaussians). |
| -sel<em>FileName</em><a | name="spr_fv_gauss.sel" class="el"> The input select file name (tied gaussian). |
| -am_opt<em>Options</em><a | name="spr_fv_gauss.am_opt" class="el"> Extra options for loading the acoustic model. A non-default acoustic model can be selected by having '=<am_type>;' as first option. See cwr_am_tbl.c for a list of acoustic models available. |
| -top_n<em>Number(s)</em><a | name="spr_fv_gauss.top_n" class="el"> Only take the top-N gaussians into account when calculating output probabilities. If one value is given, it is used for all mixtures. Else a value per mixture must be given, separated by commas. Use '0' to set top_n to the number of gaussians in the mixture. |
| -rmg<em>rmg_params</em><a | name="spr_fv_gauss.rmg" class="el"> The parameters for the quick selection of gaussians. If one value is given, it is used for all mixtures. Else a value per mixture must be given, separated by commas. Use 'no' if no quick selection is wanted. See rm_gauss.c for a description of the parameters. |
| -LMout<em>Value</em><a | name="spr_fv_gauss.LMout" class="el"> Floor the state likelihoods of an observation using a fraction of the unconditional likelihood of the observation (weighted sum of the state likelihoods). Practically necessary if only few gaussians are evalutated (-top_n or -rmg options). The value given offset an automatically determined log10(fraction). Use -100 to turn the flooring off, and 0.0 to use the default.automatically. |
| -NOGSflag | no gauss sel Forgo the (sentence level) lexicon based Gaussian selection. The lexicon based Gaussian selection speeds up the decoding but may interfere with score normalization techniques that assume all Gaussians were evaluated. |
| -ci<em>unit | description files The two units description files seperated by white-space. The first file just lists the units (phones). The second file describes the context dependencies. |
| -cd<em>unit | description files The two units description files seperated by white-space. The first file just lists the units (phones). The second file describes the context dependencies. |
| -d<em>dictionary</em><a | name="spr_fv_gauss.d" class="el"> Dictionary file name. |
| -unwind<em>unwind | format Define the parameters to modify the word transcriptions. See spr_cwr_lex_desc_read() for more details. |
Collect the necessary statistics for fMLLR(+HLDA) and calculate the transformation matrices.
Calculate the statistics for feature space maximum likelihood linear regression (fMLLR), eventually combined with a heteroscedastic linear discriminant transformation and estimate the optimal speaker specific tranform.
1.8.6