hsmooth.tex

隐马尔可夫模型源代码

%/* ----------------------------------------------------------- */
%/*                                                             */
%/*                          ___                                */
%/*                       |_| | |_/   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 Dave Ollason  11/11/95
%


\newpage
\mysect{HSmooth}{HSmooth}

\mysubsect{Function}{HSmooth-Function}

\index{hsmooth@\htool{HSmooth}|(}
This program is used to smooth a set of context-dependent tied mixture
or discrete HMM's using deleted interpolation.  
The program operates as a replacement for the second pass of \htool{HERest}
when working in {\it parallel mode}\footnote{It is not,of course,
necessary to have multiple processors to use this program since each 
`parallel' activation can be executed sequentially on a single
processor}.  It reads in the $N$ sets of accumulator files containing the
statistics accumulated during the first pass and then interpolates the
mixture weights between each context dependent model and its corresponding 
context independent version.  The interpolation weights are chosen to 
maximise the likelihood of each deleted block with respect to the 
probability distributions estimated from all other blocks.

\mysubsect{Use}{HSmooth-Use}

\htool{HSmooth} is invoked via the command line
\begin{verbatim}
   HSmooth [options] hmmList accFile ...
\end{verbatim}
where {\tt hmmList} contains the list of context dependent models
to be smoothed and each {\tt accFile} is a file of the form
{\tt HERN.acc} dumped by a previous execution of \htool{HERest} with the {\tt -p}
option set to {\tt N}.  The HMM definitions are loaded and then for
every state and stream of every context dependent model X, the optimal
interpolation weight is found for smoothing between the mixture
weights determined from the X accumulators alone and those determined
from the context independent version of X.  The latter is computed simply
by summing the accumulators across all context dependent allophones of X.

The detailed operation of \htool{HSmooth} is controlled by the following
command line options
\begin{optlist}
  \ttitem{-b f}  Set the value of epsilon for convergence 
      in the binary chop  optimisation procedure to {\tt f}.
       The binary chop  optimisation procedure for each
       interpolation weight terminates when the gradient is
       within epsilon of zero (default 0.001).
  
  \ttitem{-c N}  Set maximum number of interpolation iterations
       for the binary chop  optimisation procedure to be N
       (default 16).

\ttitem{-d dir} Normally \htool{HSmooth} expects to find the HMM definitions
      in the current directory.  This option tells \htool{HSmooth} to look in
      the directory {\tt dir} to find them.


   \ttitem{-m N}  Set the minimum number of training examples 
    required for any model to {\tt N}.  If the actual number
    falls below this value, the HMM is not updated and the original
    parameters are used for the new version (default value 1).

  \ttitem{-o ext}  This causes the file name extensions of the
      original models (if any) to be replaced by {\tt ext}.

  \ttitem{-s file} This causes statistics on occupation of each
      state to be output to the named file.  
      
  \ttitem{-u flags} By default, \htool{HSmooth} updates all of the HMM parameters,
      that is, means, variances and transition probabilies. This 
      option causes just the parameters indicated by the {\tt flags}
      argument to be updated, this argument is a string containing one
      or more of the letters {\tt m} (mean), {\tt v} (variance) ,
      {\tt t} (transition) and {\tt w} (mixture weight).  The 
      presence of a letter enables
      the updating of the corresponding parameter set.

  \ttitem{-v f}  This sets the minimum variance (i.e. diagonal element of
      the covariance matrix) to the real value {\tt f} (default value
      0.0).
  
  \ttitem{-w f}  Any mixture weight which falls below the global
            constant {\tt MINMIX} is treated as being zero.
      When this parameter is  set,  all mixture weights  are floored
      to {\tt f * MINMIX}.
     
  \ttitem{-x ext}  By default, \htool{HSmooth} expects a HMM definition for 
      the model X to be stored in a file called {\tt X}.  This
      option causes \htool{HSmooth} to look for the HMM definition in the
      file {\tt X.ext}.

\stdoptB
\stdoptH
\stdoptM

\end{optlist}
\stdopts{HSmooth}

\mysubsect{Tracing}{HSmooth-Tracing}

\htool{HSmooth} supports the following trace options where each
trace flag is given using an octal base
\begin{optlist}
   \ttitem{00001} basic progress reporting.
   \ttitem{00002} show interpolation weights.
   \ttitem{00004} give details of optimisation algorithm.
\end{optlist}
Trace flags are set using the \texttt{-T} option or the  \texttt{TRACE} 
configuration variable.
\index{hsmooth@\htool{HSmooth}|)}


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