hvite.tex

该压缩包为最新版htk的源代码,htk是现在比较流行的语音处理软件,请有兴趣的朋友下载使用

%/* ----------------------------------------------------------- */
%/*                                                             */
%/*                          ___                                */
%/*                       |_| | |_/   SPEECH                    */
%/*                       | | | | \   RECOGNITION               */
%/*                       =========   SOFTWARE                  */ 
%/*                                                             */
%/*                                                             */
%/* ----------------------------------------------------------- */
%/*         Copyright: Microsoft Corporation                    */
%/*          1995-2000 Redmond, Washington USA                  */
%/*                    http://www.microsoft.com                */
%/*                                                             */
%/*   Use of this software is governed by a License Agreement   */
%/*    ** See the file License for the Conditions of Use  **    */
%/*    **     This banner notice must not be removed      **    */
%/*                                                             */
%/* ----------------------------------------------------------- */
%
% HTKBook - Steve Young and Julian Odell - 24/11/97
%

\newpage
\mysect{HVite}{HVite}

\mysubsect{Function}{HVite-Function}

\index{hvite@\htool{HVite}|(}
\htool{HVite} is a general-purpose Viterbi word recogniser.  It will 
match a speech file against a network of HMMs and output a transcription 
for each.  When performing N-best recognition a word level lattice 
containing multiple hypotheses can also be produced.

Either a word level lattice or a label file is read in and then
expanded using the supplied dictionary to create a model based 
network.  This allows arbitrary finite state word networks and
simple forced alignment to be specified.

This expansion can be used to create context independent, word internal
context dependent and cross word context dependent networks.  The way
in which the expansion is performed is determined automatically from
the dictionary and HMMList.  When all labels appearing in the
dictionary are defined in the HMMList no expansion of model names is 
performed.  Otherwise if all the labels in the dictionary can be 
satisfied by models dependent only upon word internal context these
will be used else cross word context expansion will be performed.
These defaults can be overridden by \htool{HNet} configuration parameters.

\htool{HVite} supports shared parameters and appropriately pre-computes 
output probabilities. 
For increased processing speed, \htool{HVite} can optionally perform a beam
search controlled by a user specified threshold (see \texttt{-t} option).
When fully tied mixture models are used, observation pruning is also provided
(see the  \texttt{-c} option).
Speaker adaptation is also supported by \htool{HVite} both in terms of 
recognition using an adapted model set or a TMF (see the \texttt{-k} option),  
and in the estimation of a transform by unsupervised adaptation using 
linear transformation  in an incremental mode (see the \texttt{-j} option) or 
in a batch mode (\texttt{-K} option).

\mysubsect{Use}{HVite-Use}

\htool{HVite} is invoked via the command line
\begin{verbatim}
   HVite [options] dictFile hmmList testFiles ...
\end{verbatim}
HVite will then either load a single network file and match this
against each of the test files \texttt{-w netFile}, or create a
new network for each test file either from the corresponding 
label file \texttt{-a} or from a word lattice \texttt{-w}.
When a new network is created for each test file the path name
of the label (or lattice) file to load is determined from the
test file name and the \texttt{-L} and \texttt{-X} options
described below.

If no \texttt{testFiles} are specified the \texttt{-w s} option must
be specified and recognition will be performed from direct audio.

The \texttt{hmmList} should contain a list of the models required to
construct the network from the word level representation.

The recogniser output is written in the form of a label file whose
path name is determined from the test file name and the \texttt{-l} and 
\texttt{-x} options described below. The list of test files can be stored 
in a script file if required.

When performing N-best recognition (see \texttt{-n N} option described
below) the output label file can contain multiple alternatives
\texttt{-n N M} and a lattice file containing multiple hypotheses can
be produced.

The detailed operation of \htool{HVite} is controlled by the following
command line options
\begin{optlist}

  \ttitem{-a} Perform alignment.  \htool{HVite} will load a label file and
        create an alignment network for each test file.

  \ttitem{-b s} Use \texttt{s} as the sentence boundary during alignment.
  
  \ttitem{-c f} Set the tied-mixture observation pruning threshold to \texttt{f}.
        When all mixtures of all models are tied to create a full
        tied-mixture system, the calculation of output probabilities
        is treated as a special case.  Only those mixture 
        component probabilities which fall within \texttt{f} of
        the maximum mixture component probability are used in calculating
        the state output probabilities (default 10.0).

  \ttitem{-d dir} This specifies the directory to search for the
        HMM definition files corresponding to the labels used in
        the recognition network.
  \ttitem{-e} When using direct audio input, output transcriptions
        are not normally saved.  When this option is set, each
        output transcription is written to a file called \texttt{PnS}
        where \texttt{n} is an integer which increments with each output
        file, \texttt{P} and \texttt{S} are strings which are by default
        empty but can be set using the configuration variables
        \texttt{RECOUTPREFIX} and \texttt{RECOUTSUFFIX}.

\ttitem{-f} During recognition keep track of full state alignment. The output
        label file will contain multiple levels. The first level will be the 
        state number and the second will be the word name (not the output symbol).

  \ttitem{-g} When using direct audio input, this option enables audio
        replay of each input utterance after it has been recognised.

  \ttitem{-h mask} Set the mask for determining which transform names are 
	to be used for the input transforms. 

  \ttitem{-i s} Output transcriptions to MLF \texttt{s}.

  \ttitem{-j i} Perform incremental MLLR adaptation every i utterances      

  \ttitem{-k} Use an input transform (default off).

  \ttitem{-l dir} This specifies the directory to store the  output label 
        files.  If this option is not used then \htool{HVite} will store 
        the label files in the same directory as the data.
      When output is directed to an MLF, this option can be used to
      add a path to each output file name.  In particular, setting the option
      \verb+-l '*'+ will cause a label file named \texttt{xxx} to be prefixed
      by the pattern \verb+"*/xxx"+ in the output MLF file.  This is useful
      for generating MLFs which are independent of the location of the 
      corresponding data files.

  \ttitem{-m} During recognition keep track of model boundaries. The output
        label file will contain multiple levels. The first level will be the 
        model number and the second will be the word name (not the output 
        symbol).


  \ttitem{-n i [N]} Use \texttt{i} tokens in each state to perform
        N-best recognition.  The number of alternative output
        hypotheses \texttt{N} defaults to 1.

  \ttitem{-o s} Choose how the output labels should be formatted.
        \texttt{s} is a string with certain letters (from \texttt{NSCTWM})
        indicating binary flags that control formatting options.
        \texttt{N} normalise acoustic scores by dividing by the duration
        (in frames) of the segment.
        \texttt{S} remove scores from output label.  By default 
        scores will be set to the total likelihood of the segment.
        \texttt{C} Set the transcription labels to start and end on
        frame centres. By default start times are set to the start
        time of the frame and end times are set to the end time of 
        the frame.
        \texttt{T} Do not include times in output label files.
        \texttt{W} Do not include words in output label files
        when performing state or model alignment.
        \texttt{M} Do not include model names in output label
        files when performing state and model alignment.

  \ttitem{-p f}  Set the word insertion log probability to \texttt{f} 
        (default 0.0).

  \ttitem{-q s} Choose how the output lattice should be formatted.
         \texttt{s} is a string with certain letters (from \texttt{ABtvaldmn})
         indicating binary flags that control formatting options.
         \texttt{A} attach word labels to arcs rather than nodes.
         \texttt{B} output lattices in binary for speed.
         \texttt{t} output node times.
         \texttt{v} output pronunciation information.
         \texttt{a} output acoustic likelihoods.
         \texttt{l} output language model likelihoods.
         \texttt{d} output word alignments (if available).
         \texttt{m} output within word alignment durations.
         \texttt{n} output within word alignment likelihoods.

  \ttitem{-r f} Set the dictionary pronunciation probability scale 
        factor to \texttt{f}. (default value 1.0).

  \ttitem{-s f} Set the grammar scale factor to \texttt{f}.
        This factor post-multiplies the language model likelihoods
        from the word lattices.  (default value 1.0).
   
  \ttitem{-t f [i l]} Enable beam searching such that any model whose 
        maximum log probability token falls more than \texttt{f} below
        the maximum for all models is deactivated.  Setting \texttt{f}
        to 0.0 disables the beam search mechanism (default value
        \texttt{0.0}). In alignment mode two extra parameters
        \texttt{i} and \texttt{l} can be specified. If the alignment
        fails at the initial pruning threshold \texttt{f}, then the
        threshold will by increased by \texttt{i} and the alignment
        will be retried. This procedure is repeated until the
        alignment succeeds or the threshold limit \texttt{l} is
        reached.        

  \ttitem{-u i} Set the maximum number of active models to \texttt{i}.
        Setting \texttt{i} to \texttt{0} disables this limit (default 0).

  \ttitem{-v f} Enable word end pruning.  Do not propagate tokens from
        word end nodes that fall more than \texttt{f} below the maximum 
        word end likelihood.  (default \texttt{0.0}).

  \ttitem{-w [s]} Perform recognition from word level networks.  If
        \texttt{s} is included then use it to define the network used
        for every file.

  \ttitem{-x ext}  This sets the extension to use for HMM definition
      files to \texttt{ext}.

  \ttitem{-y ext}  This sets the extension for output label files to
        \texttt{ext} (default \texttt{rec}).

  \ttitem{-z ext}  Enable output of lattices (if performing NBest
        recognition) with extension \texttt{ext} (default off).

  \ttitem{-L dir} This specifies the directory to find input label (when
        \texttt{-a} is specified) or network files (when \texttt{-w} is
        specified).

  \ttitem{-X s} Set the extension for the input label or network files 
        to be \texttt{s}  (default value \texttt{lab}).

\stdoptE
\stdoptG
\stdoptH
\stdoptI
\stdoptJ
\stdoptK
\stdoptP

\end{optlist}
\stdopts{HVite}

\mysubsect{Tracing}{HVite-Tracing}

\htool{HVite} supports the following trace options where each
trace flag is given using an octal base
\begin{optlist}
   \ttitem{0001} enable basic progress reporting.  
   \ttitem{0002} list observations.
   \ttitem{0004} frame-by-frame best token.
   \ttitem{0010} show memory usage at start and finish.
   \ttitem{0020} show memory usage after each utterance.
\end{optlist}
Trace flags are set using the \texttt{-T} option or the \texttt{TRACE} 
configuration variable.
\index{hvite@\htool{HVite}|)}


%%% Local Variables: 
%%% mode: latex
%%% TeX-master: "../htkbook"
%%% End: