Sophie

Sophie

distrib > Mandriva > 9.2 > i586 > by-pkgid > a804ef007a99f7d26cf24253c2994680 > files > 1400

howto-html-ja-9.1-0.5mdk.noarch.rpm

<HTML
><HEAD
><TITLE
>$B2;@<G'<1$NFbB&(B</TITLE
><META
NAME="GENERATOR"
CONTENT="Modular DocBook HTML Stylesheet Version 1.54"><LINK
REL="HOME"
TITLE="Speech Recognition HOWTO"
HREF="index.html"><LINK
REL="PREVIOUS"
TITLE="$B2;@<G'<1%=%U%H%&%'%"(B"
HREF="software.html"><LINK
REL="NEXT"
TITLE="$B=PHGJ*(B"
HREF="publications.html"></HEAD
><BODY
CLASS="SECT1"
BGCOLOR="#FFFFFF"
TEXT="#000000"
><DIV
CLASS="NAVHEADER"
><TABLE
WIDTH="100%"
BORDER="0"
CELLPADDING="0"
CELLSPACING="0"
><TR
><TH
COLSPAN="3"
ALIGN="center"
>Speech Recognition HOWTO</TH
></TR
><TR
><TD
WIDTH="10%"
ALIGN="left"
VALIGN="bottom"
><A
HREF="software.html"
>Prev</A
></TD
><TD
WIDTH="80%"
ALIGN="center"
VALIGN="bottom"
></TD
><TD
WIDTH="10%"
ALIGN="right"
VALIGN="bottom"
><A
HREF="publications.html"
>Next</A
></TD
></TR
></TABLE
><HR
ALIGN="LEFT"
WIDTH="100%"></DIV
><DIV
CLASS="SECT1"
><H1
CLASS="SECT1"
><A
NAME="INSIDE"
>6. $B2;@<G'<1$NFbB&(B</A
></H1
><DIV
CLASS="SECT2"
><H2
CLASS="SECT2"
><A
NAME="RECOGNIZERS"
>6.1. $B$I$N$h$&$KG'<1$7$F$$$k$+(B</A
></H2
><P
> 
$BG'<1%7%9%F%`$O(B 2 $B$D$N<g$JItJ,$KJ,2r$G$-$^$9(B. $B%Q%?!<%sG'<1%7%9%F%`$O(B
$BE,9g$rH=CG$9$k$?$a$K%Q%?!<%s$r4{CN$N$b$N$d3X=,$7$?%Q%?!<%s$HHf3S$7$^$9(B.
Acoustic Phonetic $B%7%9%F%`$O2;@<$NFCD'(B ($BJl2;$J$I$N2;$J$I$N2;@<(B) $B$rHf3S$9$k(B
$B$?$a$K?MBN$K4X$9$kCN<1(B ($B@<$N@8@.(B, $B$HD03P(B) $B$rMxMQ$7$^$9(B. 
$B$[$H$s$I$N8=BeE*$J%7%9%F%`$O$3$N$h$&$J%Q%?!<%sG'<1%"%W%m!<%A$K=EE@$rCV$$$F(B
$B$$$^$9(B, $B$J$<$J$i(B. $B$=$l$O8=:_$N%3%s%T%e!<%?MxMQ5;=Q$H$&$^$/7k$S$D$-(B, 
$B9b$$@:EY$rF@$d$9$$$+$i$G$9(B. </P
><P
>$B$[$H$s$I$NG'<1AuCV$O0J2<$N$h$&$JCJ3,$KJ,2r$G$-$^$9(B:</P
><P
>    <P
></P
><OL
TYPE="1"
><LI
><P
>	$B%*!<%G%#%*$N5-O?$HH/OC$N8!=P(B
        </P
></LI
><LI
><P
>        $B%W%j%U%#%k%?%j%s%0(B ($B%W%j%(%s%U%!%5%$%:(B, $B@55,2=(B, $B%P%s%G%#%s%0$J$I(B)
        </P
></LI
><LI
><P
>$B%U%l!<%_%s%0$H%&%#%s%I%&%#%s%0(B ($B%G!<%?$r;HMQ$G$-$k7A<0$KJ,2r$9$k(B)
        </P
></LI
><LI
><P
>	$B%U%#%k%?%j%s%0(B ($B99$K$=$l$>$l$N(Bwindow/frame/freq.band$B$r%U%#%k%?%j%s%0(B)
        </P
></LI
><LI
><P
>	$BHf3S$HE,9g(B ($BH/OC$NG'<1(B)
        </P
></LI
><LI
><P
>	$BF0:n(B ($BG'<1$5$l$?%Q%?!<%s$K4XO"$7$?5!G=$r<B9T(B)
        </P
></LI
></OL
></P
><P
>        
$B$=$l$>$l$NCJ3,$OC1=c$K8+$($^$9$,(B, $B0l$D0l$D$OB??t$N0[$J$k(B
($B$=$7$F$H$-$I$-$O40A4$K5U$N(B) $B5;=Q$rMxMQ$7$F$$$^$9(B.</P
><P
>(1) $B%*!<%G%#%*(B/$BH/@<$NO?2;(B: $B$$$m$$$m$JJ}K!$,$"$j$^$9(B. 
$B;O$a$O<~0O$N%*!<%G%#%*$N%l%Y%k(B ($B$$$/$D$+$N>l9g$G$O2;6A$N%(%M%k%.!<(B) $B$r(B
$BO?2;$5$l$F$$$k%5%s%W%k$HHf3S$9$k$3$H$G$9(B.  $B=*E@$NH=JL$O(B, $BOC<T$,(B
$B8F5[$d$?$aB)(B, $B;u$NLD$-(B, $B%(%3!<$J$I$N(B "artifacts"
$B$r;D$7$,$A$J$N$G$5$i$K:$Fq$G$9(B. </P
><P
>(2) $B%W%j%U%#%k%?%j%s%0(B: $BG'<1%7%9%F%`$NB>$N5!G=$K0MB8$7$F(B,
$B$$$m$$$m$JJ}K!$G9T$J$o$l$^$9(B. $B:G$b0lHLE*$JJ}K!$O(B,  
$B%5%s%W%k$N=`Hw$N$?$a$K0lO"$N%*!<%G%#%*%U%#%k%?!<$r;HMQ$9$k(B
 "Bank-of-Filters" $BK!$H(B, 
$B:90[(B($B8m:9(B)$B$N7W;;$N$?$a$KM=B,5!G=$r;HMQ$9$k(B  Linear Predictive Coding $BK!$G$9(B.
$B0[$J$k7A<0$N%9%Z%/%H%k2r@O$bMxMQ$5$l$^$9(B. </P
><P
>(3) $B%U%l!<%_%s%0(B/$B%&%$%s%I%&%#%s%0$O%5%s%W%k%G!<%?$rFCDj$NBg$-$5$KJ,3d$9$k$3$H$G$9(B. 
$B$3$l$O$7$P$7$P(B step2 $B$H(B step4 $B$X?J$_$^$9(B. $B$3$NCJ3,$OJ,@O$N$?$a$K(B
$B%5%s%W%k6-3&(B ($BJ%$N%+%A%s$H$$$&2;$J$I$r=|$/$3$H(B) $B$rMQ0U$9$k$3$H$b4^$s$G$$$^$9(B. </P
><P
>(4) $BDI2C$N%U%#%k%?%j%s%0$O$$$D$bB8:_$9$k$o$1$G$O$"$j$^$;$s(B. 
$B$3$l$OHf3S$HE,9g$NA0$G$N$=$l$>$l$N%&%#%s%I%&$KBP$9$k:G8e$N=`Hw$G$9(B. 
$B$7$P$7$P$3$l$O;~4V$NG[CV$H@55,2=$+$i9=@.$5$l$^$9(B. </P
><P
>(5)$B$NHf3S$HE,9g$K$D$$$F$O2DG=$J5;=Q$,BgNL$K$"$j$^$9(B. $B$[$H$s$I$O(B
$B8=:_$N%&%#%s%I%&$H4{CN$N%5%s%W%k$NHf3S$rI,MW$H$7$^$9(B.Hidden Markov Models (HMM), $B<~GH?t2r@O(B, $B:90[2r@O(B, $B@~7?Be?t$N5;=Q(B/$B6aF;(B, $B%9%Z%/%H%kOD$rMxMQ$9$k(B
$BJ}K!$d;~4VOD6JK!$b$"$j$^$9(B. 
$B$3$l$i$N$9$Y$F$NJ}K!$O0lCW$N3NN($H@:EY$rF@$k$?$a$K;HMQ$5$l$F$$$^$9(B. </P
><P
>(6) $BF0:n$O3+H/<T$,K>$s$@$3$H$@$1$G$9(B. </P
></DIV
><DIV
CLASS="SECT2"
><H2
CLASS="SECT2"
><A
NAME="DIGITALAUDIO"
>6.2. $B%G%#%8%?%k%*!<%G%#%*$N4pAC(B</A
></H2
><P
>$B%*!<%G%#%*$OK\<AE*$K%"%J%m%0$J8=>]$G$9(B. $B%G%#%8%?%k%5%s%W%k$G(B
$BO?2;$9$k$3$H$O(B, $B%^%$%/$+$i$N%"%J%m%0?.9f$r%5%&%s%I%+!<%I>e$N(B
A/D $B%3%s%P!<%?$G%G%#%8%?%k?.9f$KJQ49$9$k$3$H$G$9(B. $B%^%$%/$,F0:n$7$F$$$l$P(B, 
$B2;GH$O%^%$%/$NCf$N<'@P$NMWAG$r?6F0$5$;(B, $B%5%&%s%I%+!<%I$X$NEEN.(B
($B%9%T!<%+$,5U$KF0:n$7$F$$$k$H9M$($F$/$@$5$$(B) $B$rH/@8$5$;$^$9(B. $B4pK\E*$K(B
$B$=$N(BA/D$B%3%s%P!<%?$OFCDj$N4V3V$G$NEE05$NCM$r5-O?$7$^$9(B. </P
><P
>$B$3$N2aDx$NCf$K(B2$B$D$N=EMW$JMWAG$,$"$j$^$9(B. 1$B$D$a$O(B "sample rate",  $B$"$k$$$O(B
 $B$I$N$h$&$JIQEY$GEE05$r5-O?$9$k$N$+$H$$$&$b$N(B. 2$B$D$a$O(B "bits per second", 
$B$I$N$h$&$J@:EY$GCM$,5-O?$5$l$k$+$H$$$&$b$N$G$9(B. 3$B$D$a$NMWAG$O%A%c%s%M%k$N?t(B
($B%b%N$+%9%F%l%*$+(B), $B$7$+$7(B, $BB?$/$N(B ASR $B%"%W%j%1!<%7%g%s$G$O%b%N$G==J,$G$9(B. 
$BB?$/$N%"%W%j%1!<%7%g%s$G$O$3$l$i$N%Q%i%a!<%?$KM=$a@_Dj$5$l$?CM$r(B
$B;HMQ$7$F(B, $B%f!<%6$OJ8=q$K=q$+$l$F$$$J$$$+$.$jJQ99$9$k$Y$-$G$O$"$j$^$;$s(B. 
$B3+H/<T$O0[$J$kCM$G$=$N%"%k%4%j%:%`$H$O2?$,9g$&$N$+$r(B
$B<B83$9$k$3$H$G7hDj$9$k$Y$-$G$9(B. </P
><P
>$B$=$l$G$O(B, ASR $B$K$D$$$F$O$I$N$h$&$J%5%s%W%k%l!<%H$,NI$$$N$G$7$g$&$+(B?
$B2;@<$OHf3SE*Dc$$BS0hI}(B ($B$[$H$s$I(B 100Hz $B$+$i(B 8kHz) $B$G$9$+$i(B, 8000 samples/sec
 (8kHz) $B$O$[$H$s$I$N4pK\E*$J(B ASR $B$KBP$7$F$O==J,$G$9(B. $B$7$+$7(B, $B$5$i$K@53N$J(B
$B9b<~GH?t$N>pJs$rF@$i$l$k$N$G(B 16000 samples/sec(16kHz) $B$r9%$`?M$b$$$^$9(B. 
$B$b$7=hM}G=NO$,$"$l$P(B 16kHz $B$r;H$&$Y$-$G$9(B. $B$[$H$s$I$N(B ASR $B%"%W%j%1!<%7%g%s$G$O(B
22kHz $B0J>e$N%5%s%W%j%s%0%l!<%H$OL5BL$G$9(B. </P
><P
>$B$=$7$F(B $B$I$N$h$&$JCM$,(B "bits per sample" (1$B%5%s%W%kEv$j$N%S%C%H?t(B) $B$K$D$$$FNI$$$N$G$7$g$&$+(B?
 8 bits per sample $B$O(B 0 $B$+$i(B 255 $B$N4V$GCM$r5-O?$7$^$9(B, $B$3$l$O%^%$%/@.J,$NBg$-$5$,(B
256$B8D$NCf$N(B1$B$D$G$"$k$H$$$&$3$H$r0UL#$7$^$9(B. 16 bits per sample$B$O(B
$B@.J,$NBg$-$5$r(B 65536 $B8D$KJ,3d$7$^$9(B. $B%5%s%W%k%l!<%H$bF1MM$G$9(B. $BHf3S$N$?$a$K(B, 
$B2;3ZMQ(B Compact Disc $B$O(B 44kHz $B$G(B 16 bits per sample$B$G%(%s%3!<%I$5$l$F$$$^$9(B. </P
><P
>$B;HMQ$5$l$k%(%s%3!<%G%#%s%0%U%)!<%^%C%H$O@~7?$J(B
$BId9fIU$"$k$$$OId9f$J$7@0?t$N$h$&$KC1=c$G$"$k$Y$-$G$9(B.
U-Law/A-Law $B%"%k%4%j%:%`$^$?$OB>$N05=LK!$r;HMQ$9$k(B
$B$3$H$OIaDL2ACM$,$"$j$^$;$s(B,
$B$J$<$J$i$=$l$O7W;;G=NO$N%3%9%H$,$+$+$j(B, 
$B%3%9%H$K8+9g$C$?G=NO$r==J,$KF@$k$3$H$,$G$-$J$$$+$i$G$9(B</P
></DIV
></DIV
><DIV
CLASS="NAVFOOTER"
><HR
ALIGN="LEFT"
WIDTH="100%"><TABLE
WIDTH="100%"
BORDER="0"
CELLPADDING="0"
CELLSPACING="0"
><TR
><TD
WIDTH="33%"
ALIGN="left"
VALIGN="top"
><A
HREF="software.html"
>Prev</A
></TD
><TD
WIDTH="34%"
ALIGN="center"
VALIGN="top"
><A
HREF="index.html"
>Home</A
></TD
><TD
WIDTH="33%"
ALIGN="right"
VALIGN="top"
><A
HREF="publications.html"
>Next</A
></TD
></TR
><TR
><TD
WIDTH="33%"
ALIGN="left"
VALIGN="top"
>$B2;@<G'<1%=%U%H%&%'%"(B</TD
><TD
WIDTH="34%"
ALIGN="center"
VALIGN="top"
>&nbsp;</TD
><TD
WIDTH="33%"
ALIGN="right"
VALIGN="top"
>$B=PHGJ*(B</TD
></TR
></TABLE
></DIV
></BODY
></HTML
>