openviron.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                  */ 
%/*                                                             */
%/*                                                             */
%/* ----------------------------------------------------------- */
%/* developed at:                                               */
%/*                                                             */
%/*      Speech Vision and Robotics group                       */
%/*      Cambridge University Engineering Department            */
%/*      http://svr-www.eng.cam.ac.uk/                          */
%/*                                                             */
%/*      Entropic Cambridge Research Laboratory                 */
%/*      (now part of Microsoft)                                */
%/*                                                             */
%/* ----------------------------------------------------------- */
%/*         Copyright: Microsoft Corporation                    */
%/*          1995-2000 Redmond, Washington USA                  */
%/*                    http://www.microsoft.com                 */
%/*                                                             */
%/*          2001-2002 Cambridge University                     */
%/*                    Engineering Department                   */
%/*                                                             */
%/*   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 15/11/95
%

\mychap{The Operating Environment}{openviron}

\sidepic{Tool.shell}{80}{
This chapter discusses the various ways of  controlling the operation of
\HTK\ tools along with related aspects of file system organisation, error
reporting and memory management.   All of the operating 
system\index{operating system} and user
interface functions are provided by the \HTK\ 
module \htool{HShell}. 
Memory management is a low level function which is largely invisible to
the user, but it is useful to have a basic understanding of it in order to
appreciate memory requirements and interpret diagnostic output from
tools.  Low level memory management in \HTK\ is provided by
\htool{HMem} and the management of 
higher level structures such as vectors
and matrices is provided by \htool{HMath}.
}
\index{hshell@\htool{HShell}}
\index{hmath@\htool{HMath}}
\index{hmem@\htool{HMem}}

The behaviour of a \HTK\ tool depends on three sources of information.
Firstly,
all \HTK\ tools are executed by issuing commands to the operating
system shell.  
Each command typically contains the names of the various files that the
tool needs to function and a number of optional arguments which 
control the detailed behaviour of the tool.  
Secondly, 
as noted in chapter~\ref{c:htkoview} and shown in the adjacent figure, every \HTK\ tool uses a set
of standard library modules to interface to the various file types and
to connect with the outside world.  Many of these modules can be
customised by setting parameters in a {\it configuration file}.
Thirdly, a small number of parameters are specified using
environment variables.  

Terminal output mostly depends on the specific tool 
being used, however, there
are some generic output functions which are provided by the library
modules and which are therefore common across tools.  These include
version reporting, memory usage and error reporting.

Finally, \HTK\ can read and write most data sources through pipes\index{pipes} as an alternative
to direct input and output from data files.  This allows 
filters to be used, and in particular, it allows many of the external
files used by \HTK\ to be stored directly in compressed form and then decompressed
\textit{on-the-fly} when the data is read back in.

All of the above is discussed in more detail in the following sections.

\mysect{The Command Line}{cmdline}

The general form of command line for invoking a tool 
is\footnote{All of the examples in this book assume the
UNIX Operating System and the C Shell but the principles apply to
any OS which supports hierarchical files and command line arguments}
\index{command line!arguments}\index{command line!options}
\begin{verbatim}
           tool [options] files ...
\end{verbatim}
Options always consist of a dash followed by a single letter.  Some options
are followed by an argument as follows
\begin{tabbing}
+++++ \= ++++++++++++ \=  \kill
\>           \texttt{-i} \>  - a switch option \\
\>           \texttt{-t 3} \> - an integer valued option \\
\>           \texttt{-a 0.01} \> - a float valued option \\
\>           \texttt{-s hello} \> - a string valued option 
\end{tabbing}   
Option names consisting of a capital letter are common across all tools
(see section~\ref{s:stdopts}).
Integer arguments\index{command line!integer argument formats} may be given 
in any of the standard C formats, for example,
\texttt{13}, \texttt{0xD} and \texttt{015} all represent the same number.
Typing the name of a tool on its own always causes a short summary of
the command line options to be printed in place of its normal
operation.  For example, typing
\begin{verbatim}
    HERest
\end{verbatim}
would result in the following output
\begin{verbatim}
    USAGE: HERest [options] hmmList dataFiles...

     Option                                   Default

     -c f    Mixture pruning threshold          10.0
     -d s    dir to find hmm definitions       current
     -m N    set min examples needed per model   3
     -o s    extension for new hmm files        as src
     -p N    set parallel mode to N             off
     ...
\end{verbatim}
The first line shows the names of the required files and the rest
consists of a listing of each option, its meaning, and its default
value.

The precise naming convention for specifying files depends on the operating
system being used, but \HTK\ always assumes the existence of a hierarchical
file system and it maintains a distinction between directory paths
and file names.

In general, a file will be located either in
the current directory, some subdirectory of the current directory or some
subdirectory of the root directory.  For example, in the command
\begin{verbatim}
      HList s1 dir/s2 /users/sjy/speech/s3
\end{verbatim}
file \texttt{s1} must be in the current directory, \texttt{s2} must be in the
directory \texttt{dir} within the current directory and \texttt{s3} must be in the
directory \texttt{/users/sjy/speech}.

Some tools allow directories to be specified via configuration
parameters and command line options.  In all cases, the final path
character (eg / in UNIX) need not (but may be)
included.  For example, both of the following are acceptable and have
equivalent effect
\begin{verbatim}
      HInit -L mymodels/new/  hmmfile data*
      HInit -L mymodels/new   hmmfile data*
\end{verbatim}  
where the \texttt{-L} option specifies the directory in which to find the 
label files associated with the data files. 

\mysect{Script Files}{script}

Tools which require a potentially very long list of files (e.g.\ training
tools) always allow the files to be specified in a script\index{files!script}
file\index{script files} via the \texttt{-S} option instead of via the command
line.
\index{standard options!aaas@\texttt{-S}}This is particularly useful when running
under an OS with limited file name expansion capability.  Thus, for example,
\htool{HInit}\index{hinit@\htool{HInit}} may be invoked by either
\begin{verbatim}
     HInit hmmfile s1 s2 s3 s4 s5 ....
\end{verbatim}  
or
\begin{verbatim}
     HInit -S filelist hmmfile
\end{verbatim}          
where \texttt{filelist} holds the list of files \texttt{s1}, \texttt{s2}, etc.
Each file listed in a script should be separated by white space or a new line.
Usually, files are listed on separate lines, however, when using
\htool{HCopy}\index{hcopy@\htool{HCopy}} which read pairs of files as its
arguments, it is normal to write each pair on a single line.  Script files
should only be used for storing ellipsed\index{command line!ellipsed arguments}
file list arguments. Note that shell meta-characters should not be
used in script files and will not be interpreted by the \HTK\ tools.

Starting with \HTK\ 3.1 the syntax of script fies has been extended.
In addition to directly specifying the name of a physical file it is
possible to define aliases and to select a segment from a file. The
general syntax of an extended filename \index{extended filenames} is
\begin{verbatim}
     logfile=physfile[s,e]   
\end{verbatim}
where \texttt{logfile} is the logical filename used by the HTK tools
and will appear in mlf files and similar. \texttt{physfile} is the
physical name of the actual file on disk that will be accessed and
\texttt{s} and \texttt{e} are indices that can be used to select only
a segment of the file. One example of a use of this feature is the
evaluation of different segmentations of the audio data. A new
segmentation can be used by creating a new script file without having
to create multiple copies of the data.

A typical script file might look like:
\begin{verbatim}
s23-0001-A_000143_000291.plp=/data/plp/complete/s23-0001-A.plp[143,291]
s23-0001-A_000291_000500.plp=/data/plp/complete/s23-0001-A.plp[291,500]
s23-0001-A_000500_000889.plp=/data/plp/complete/s23-0001-A.plp[500,889]
\end{verbatim}


\mysect{Configuration Files}{config}

Configuration files\index{configuration files|(}\index{files!configuration} 
are used for customising the \HTK\ working environment.
They consist of a list of parameter-values pairs along with an optional
prefix which limits the scope of the parameter to a specific module
or tool.

The name of a configuration file can be specified explicitly on the
command line using the \texttt{-C}\index{standard
options!aaac@\texttt{-C}} command.  For example, when executing
\begin{verbatim}
    HERest ... -C myconfig  s1 s2 s3 s4 ...
\end{verbatim}
The operation of \htool{HERest} will depend on the parameter settings
in the file \texttt{myconfig}.    

When an explicit configuration file is specified, only those parameters mentioned
in that file are actually changed and all other parameters retain their
default values.  These defaults are built-in.  However, user-defined defaults
can be set by assigning the name of a default configuration file to
the environment variable \texttt{HCONFIG}\index{HCONFIG@\texttt{HCONFIG}}.
\index{configuration files!default}
Thus, for example, using the UNIX C Shell, writing
\begin{verbatim}
    setenv HCONFIG myconfig
    HERest ... s1 s2 s3 s4 ...
\end{verbatim}
would have an identical effect to the preceding example.  However, in this
case, a further refinement of the configuration values is possible since
the opportunity to specify an explicit configuration file on the command
line remains.  For example, in
\begin{verbatim}
    setenv HCONFIG myconfig
    HERest ... -C xconfig  s1 s2 s3 s4 ...
\end{verbatim}

\sidefig{Config}{60}{Defining a Configuration}{2}
\noindent
the parameter values in \texttt{xconfig} will over-ride those in 
\texttt{myconfig} which in turn will over-ride the built-in defaults.
In practice, most \HTK\ users will set general-purpose
default configuration values using \texttt{HCONFIG} and will then over-ride
these as required for specific tasks using the \texttt{-C} command line option.
This is illustrated in Fig.~\href{f:Config} where the darkened rectangles
indicate \textit{active} parameter definitions.
Viewed from above,