.#config.tex.1.6

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 - Dave Ollason and Steve Young 11/11/95
%
% Language modelling details added by Gareth Moore 15-16/01/02

\mychap{Configuration Variables}{confvars}

\index{configuration variables!summary}
This chapter tabulates all configuration variables used in \HTK.

\mysect{Configuration Variables used in Library Modules}{coninlib}

%\begin{table}[h]
\begin{center}
\tablecaption{Library Module Configuration Variables}
\tablehead{\hline
Module & Name & Default & Description  \\ \hline\hline}
\tabletail{\hline}
\begin{supertabular}{|p{1.4cm}|l|l|p{6.6cm}|}

% all
  & \texttt{TRACE} & \texttt{0} & Trace setting\\ \hline

% HParm HWave
\htool{HParm} & \texttt{SOURCEFORMAT} & \texttt{HTK}
  & File format of source \\ \cline{2-4}
\htool{HWave} & \texttt{TARGETFORMAT} & \texttt{HTK} & File format of target \\ \hline

% HLabel HAudio HWave HParm
\htool{HLabel} \htool{HAudio} \htool{HWave} \htool{HParm} 
  & \texttt{SOURCERATE} & \texttt{0.0} & Sample rate of source in 100ns units \\ \hline

% HParm
\htool{HParm} \htool{HWave} 
  & \texttt{TARGETRATE} & \texttt{0.0} & Sample rate of target in 100ns units \\ \hline

% HAudio
  & \texttt{LINEOUT} & \texttt{T}   & Enable audio output to machine line output \\ \cline{2-4}
  & \texttt{PHONESOUT} & \texttt{T} & Enable audio output to machine phones output \\ \cline{2-4}
\htool{HAudio}
  & \texttt{SPEAKEROUT} & \texttt{F}& Enable audio output to machine internal speaker \\ \cline{2-4}
  & \texttt{LINEIN} & \texttt{T} & Enable audio input from machine line input \\ \cline{2-4}
  & \texttt{MICIN}  & \texttt{F} & Enable audio input from machine mic input \\ \hline

% HWave
  & \texttt{NSAMPLES} &  & Num samples in alien file input via a pipe\\ \cline{2-4}
\htool{HWave} & \texttt{HEADERSIZE}  &   & Size of header in an alien file\\ \cline{2-4}
  & \texttt{BYTEORDER} &   & Define byte order \texttt{VAX} or other\\ \cline{2-4}
  & \texttt{STEREOMODE} &   & Select channel: \texttt{RIGHT} or \texttt{LEFT} \\ \hline

% HParm
  & \texttt{SOURCEKIND} & \texttt{ANON} & Parameter kind of source \\ \cline{2-4}
  & \texttt{TARGETKIND} & \texttt{ANON} & Parameter kind of target \\ \cline{2-4}
  & \texttt{MATTRANFN } &  & Input transformation file  \\ \cline{2-4}
  & \texttt{SAVECOMPRESSED} & \texttt{F} & Save the output file in compressed form \\ \cline{2-4}
  & \texttt{SAVEWITHCRC} & \texttt{T} & Attach a checksum to output parameter file \\ \cline{2-4}
\htool{HParm}  
  & \texttt{ADDDITHER} & \texttt{0.0} & Level of noise added to input signal \\ \cline{2-4} 
  & \texttt{ZMEANSOURCE} & \texttt{F} & Zero mean source waveform before analysis \\ \cline{2-4}
  & \texttt{WINDOWSIZE} & \texttt{256000.0} & Analysis window size in 100ns units \\ \cline{2-4}
  & \texttt{USEHAMMING} & \texttt{T} & Use a Hamming window \\ \cline{2-4}
  & \texttt{DOUBLEFFT} & \texttt{F} & Use twice the required size for FFT \\ \cline{2-4}
  & \texttt{PREEMCOEF} & \texttt{0.97} & Set pre-emphasis coefficient \\ \cline{2-4}
  & \texttt{LPCORDER} &  \texttt{12} &  Order of lpc analysis \\ \cline{2-4}
  & \texttt{NUMCHANS} & \texttt{20} & Number of filterbank channels \\ \cline{2-4}
  & \texttt{LOFREQ} & \texttt{-1.0} & Low frequency cut-off in fbank analysis \\ \cline{2-4}
  & \texttt{HIFREQ}  & \texttt{-1.0} & High frequency cut-off in fbank analysis \\  \cline{2-4}
  & \texttt{WARPFREQ} & \texttt{1.0} & Frequency warping factor \\ \cline{2-4}
  & \texttt{WARPLCUTOFF} &  & Lower frequency threshold for non-linear warping  \\ \cline{2-4}
  & \texttt{CMEANDIR} &  & Directory to find cepstral mean vecotrs  \\ \cline{2-4}
  & \texttt{CMEANMASK} &  & Filename mask for cepstral mean vectors  \\ \cline{2-4}
  & \texttt{CMEANPATHMASK} &  & Path name mask for cepstral mean vectors, the matched string is used to extend CMEANDIR string \\ \cline{2-4}
  & \texttt{VARSCALEDIR} &  & Directory to find cepstral variance vectors  \\ \cline{2-4}
  & \texttt{VARSCALEMASK} &  & Filename mask for cepstral variance vectors  \\ \cline{2-4}
  & \texttt{VARSCALEPATHMASK} &  & Path name mask for cepstral variance vectors, the matched string is used to extend VARSCALEDIR string\\ \cline{2-4}
  & \texttt{VARSCALEFN} &  & Filename of global variance scaling vector \\ \cline{2-4}
  & \texttt{COMPRESSFACT} & 0.33 & Amplitude compression factor for PLP \\ \hline

% HLabel HParm
\htool{HLabel}  \htool{HParm} 
  & \texttt{V1COMPAT} & \texttt{F} & HTK V1 compatibility setting \\ \hline

%\end{tabular}
%\end{center}
%\caption{Library Module Configuration Variables}
%\end{table}
%\clearpage

%\begin{table}[h]
%\begin{center}
%\begin{tabular}{|p{1.4cm}|p{3.0cm}|p{1.1cm}|p{6.6cm}|} \hline
%Module & Name & Default & Description  \\ \hline\hline

% HWave HShell
\htool{HWave} 
  & \texttt{NATURALREADORDER} & \texttt{F} & Enable natural read order for binary files \\ \cline{2-4} 
\htool{HShell} 
  & \texttt{NATURALWRITEORDER} & \texttt{F} & Enable natural write order for binary files \\ \hline

% HParm
  & \texttt{USEPOWER}  & \texttt{F} & Use power not magnitude in fbank analysis \\ \cline{2-4}
  & \texttt{NUMCEPS}   & \texttt{12} & Number of cepstral parameters \\  \cline{2-4}
  & \texttt{CEPLIFTER} &   \texttt{22} & Cepstral liftering coefficient \\ \cline{2-4}
  & \texttt{ENORMALISE} & \texttt{T} & Normalise log energy \\  \cline{2-4}
  & \texttt{ESCALE} & \texttt{0.1} & Scale log energy \\  \cline{2-4} 
  & \texttt{SILFLOOR} & \texttt{50.0} & Energy silence floor in dBs \\  \cline{2-4} 
  & \texttt{DELTAWINDOW} & \texttt{2} & Delta window size \\ \cline{2-4} 
  & \texttt{ACCWINDOW} & \texttt{2} & Acceleration window size \\ \cline{2-4} 
  & \texttt{VQTABLE} & \texttt{NULL} & Name of VQ table \\  \cline{2-4}
  & \texttt{SIMPLEDIFFS} & \texttt{F} & Use simple differences for delta calculations \\ \cline{2-4}
  & \texttt{RAWENERGY}  & \texttt{T} & Use raw energy \\ \cline{2-4}
  & \texttt{AUDIOSIG} & \texttt{0} & Audio signal number for remote control \\ \cline{2-4}
  & \texttt{USESILDET}  & \texttt{F} & Enable speech/silence detector \\ \cline{2-4}
  & \texttt{MEASURESIL} & \texttt{T} & Measure background silence level \\ \cline{2-4}
\htool{HParm} 
  & \texttt{OUTSILWARN} & \texttt{T} & Print a warning message to {\tt stdout} before 
measuring audio levels \\ \cline{2-4}
  & \texttt{SPEECHTHRESH} & \texttt{9.0} & Threshold for speech above silence level (in dB) \\ \cline{2-4}
  & \texttt{SILENERGY}    & \texttt{0.0} & Average background noise level (in dB) - will 
normally be measured rather than supplied in configuration \\ \cline{2-4}
  & \texttt{SPCSEQCOUNT}  & \texttt{10} & Window over which speech/silence decision reached \\ \cline{2-4}
  & \texttt{SPCGLCHCOUNT} & \texttt{0} & Maximum number of frames marked as silence in window which is
  classified as speech whilst expecting start of speech \\ \cline{2-4}  
  & \texttt{SILSEQCOUNT}  & \texttt{100} & Number of frames classified as silence needed to mark end of
  utterance \\ \cline{2-4}
  & \texttt{SILGLCHCOUNT} & \texttt{2} & Maximum number of frames marked as silence in window which is
  classified as speech whilst expecting silence \\ \cline{2-4}  
  & \texttt{SILMARGIN} & \texttt{40} & Number of extra frames included before and after start and end of
  speech marks from the speech/silence detector \\ \hline

% HLabel
  & \texttt{STRIPTRIPHONES} & \texttt{F} & Enable triphone stripping \\ \cline{2-4}
  & \texttt{TRANSALT} & \texttt{0} & Filter all but specified label alternative \\ \cline{2-4}
\htool{HLabel} 
  & \texttt{TRANSLEV} & \texttt{0} & Filter all but specified label level \\ \cline{2-4}
  & \texttt{LABELSQUOTE} & \texttt{NULL} & Select method for quoting in label files \\ \cline{2-4}
  & \texttt{SOURCELABEL} & \texttt{HTK} & Source label format \\ \cline{2-4}
  & \texttt{TARGETLABEL} & \texttt{HTK} & Target label format \\ \hline

%\end{tabular}
%\end{center}
%\caption{Library Module Configuration Variables (cont)}
%\end{table}

%\begin{table}[h]
%\begin{center}
%\begin{tabular}{|p{1.4cm}|p{2.6cm}|p{1.5cm}|p{6.6cm}|} \hline
%Module & Name & Default & Description  \\ \hline\hline

% HMem
\htool{HMem} & \texttt{PROTECTSTAKS} & \texttt{F} & Enable stack protection \\ \hline


% HModel
  & \texttt{CHKHMMDEFS} & \texttt{T} & Check consistency of HMM defs \\ \cline{2-4}
  & \texttt{SAVEBINARY} & \texttt{F} & Save HMM defs in binary format \\ \cline{2-4}
  & \texttt{KEEPDISTINCT} & \texttt{F} & Keep orphan HMMs in distinct files \\ \cline{2-4}
\htool{HModel} 
  & \texttt{SAVEGLOBOPTS} & \texttt{T} & Save \hmmt{o} with HMM defs \\ \cline{2-4}
  & \texttt{SAVEREGTREE} & \texttt{F} & Save \hmmt{r} macros with HMM defs \\ \cline{2-4}
  & \texttt{SAVEBASECLASS} & \texttt{F} & Save \hmmt{b} macros with HMM defs \\ \cline{2-4}
  & \texttt{SAVEINPUTXFORM} & \texttt{T} & Save \hmmt{i} macros with HMM defs \\ \cline{2-4}
  & \texttt{ORPHANMACFILE} & \texttt{NULL} & Last resort file for new macros \\ \cline{2-4}
  & \texttt{HMMSETKIND} & \texttt{NULL} & Kind of HMM Set \\ \cline{2-4}
  & \texttt{ALLOWOTHERHMMS} & \texttt{T} & Allow MMFs to contain HMM definitions which are 
  not listed in the HMM List \\ \cline{2-4}
  & \texttt{DISCRETELZERO}  & \texttt{F} & Map DLOGZERO to LZERO in output probability 
  calculations \\ \hline

% HNet
  & \texttt{FORCECXTEXP} & \texttt{F} & Force triphone context expansion to get 
  model names (is overridden by \texttt{ALLOWCXTEXP}) \\ \cline{2-4}
  & \texttt{FORCELEFTBI} & \texttt{F} & Force left biphone
  context expansion to get model names ie. don't try triphone names \\ \cline{2-4}
  & \texttt{FORCERIGHTBI} & \texttt{F} & Force right biphone
  context expansion to get model names ie. don't try triphone names \\ \cline{2-4}