[xiph-cvs] cvs commit: speex/doc manual.lyx
Jean-Marc Valin
jm at xiph.org
Thu Oct 31 16:02:57 PST 2002
jm 02/10/31 19:02:57
Modified: . configure.in
doc manual.lyx
Log:
...
Revision Changes Path
1.35 +1 -1 speex/configure.in
Index: configure.in
===================================================================
RCS file: /usr/local/cvsroot/speex/configure.in,v
retrieving revision 1.34
retrieving revision 1.35
diff -u -r1.34 -r1.35
--- configure.in 26 Oct 2002 04:51:10 -0000 1.34
+++ configure.in 1 Nov 2002 00:02:57 -0000 1.35
@@ -5,7 +5,7 @@
SPEEX_MAJOR_VERSION=0
SPEEX_MINOR_VERSION=99
SPEEX_MICRO_VERSION=0
-SPEEX_VERSION=1.0beta2
+SPEEX_VERSION=1.0beta3-cvs
SPEEX_BINARY_AGE=0
SPEEX_INTERFACE_AGE=0
<p><p>1.31 +309 -66 speex/doc/manual.lyx
Index: manual.lyx
===================================================================
RCS file: /usr/local/cvsroot/speex/doc/manual.lyx,v
retrieving revision 1.30
retrieving revision 1.31
diff -u -r1.30 -r1.31
--- manual.lyx 26 Oct 2002 04:51:10 -0000 1.30
+++ manual.lyx 1 Nov 2002 00:02:57 -0000 1.31
@@ -29,7 +29,7 @@
The Speex Codec Manual
\newline
-(draft for Speex 1.0beta2)
+(draft for Speex 1.0beta3)
\layout Author
Jean-Marc Valin
@@ -96,8 +96,12 @@
Introduction to Speex
\layout Standard
-The Speex project has been started because there was a need for a speech
- codec that was open-source and free from software patents.
+The Speex project (
+\family typewriter
+http://www.speex.org/
+\family default
+) has been started because there was a need for a speech codec that was
+ open-source and free from software patents.
These are essential conditions for being used by any open-source software.
There is already Vorbis that does general audio, but it is not really suitable
for speech.
@@ -121,8 +125,12 @@
without compromising too much on the efficiency of the codec.
\layout Standard
-All this led us to the choice of CELP as the encoding technique to use for
- Speex.
+All this led us to the choice of CELP
+\begin_inset LatexCommand \index{CELP}
+
+\end_inset
+
+ as the encoding technique to use for Speex.
One of the main reasons is that CELP has long proved that it could do the
job and scale well to both low bit-rates (think DoD CELP @ 4.8 kbps) and
high bit-rates (think G.728 @ 16 kbps).
@@ -132,25 +140,65 @@
The main characteristics can be summerized as follows:
\layout Itemize
-Free software/open-source, patent and royalty-free
+Free software/open-source
+\begin_inset LatexCommand \index{open-source}
+
+\end_inset
+
+, patent
+\begin_inset LatexCommand \index{patent}
+
+\end_inset
+
+ and royalty-free
\layout Itemize
-Integration of wideband and narrowband in the same bit-stream
+Integration of wideband
+\begin_inset LatexCommand \index{wideband}
+
+\end_inset
+
+ and narrowband
+\begin_inset LatexCommand \index{narrowband}
+
+\end_inset
+
+ in the same bit-stream
\layout Itemize
Wide range of bit-rates available
\layout Itemize
-Dynamic bit-rate switching and variable bit-rate (VBR)
+Dynamic bit-rate switching and Variable Bit-Rate
+\begin_inset LatexCommand \index{variable bit-rate}
+
+\end_inset
+
+ (VBR)
\layout Itemize
-Voice Activity Detection (VAD, integrated with VBR)
+Voice Activity Detection
+\begin_inset LatexCommand \index{voice activity detection}
+
+\end_inset
+
+ (VAD, integrated with VBR)
\layout Itemize
Variable complexity
+\begin_inset LatexCommand \index{complexity}
+
+\end_inset
+
+
\layout Section
\pagebreak_top
Introduction to CELP Coding
+\begin_inset LatexCommand \index{CELP}
+
+\end_inset
+
+
\layout Standard
Speex is based on CELP, which stands for Code Excited Linear Prediction.
@@ -182,68 +230,56 @@
\layout Subsection
-Linear Prediction
-\layout Standard
-
-The linear prediction model represents each speech sample as linear combination
- of past samples, plus an error signal called the excitation (or residual).
-\begin_inset Formula \[
-x(n)=\sum _{i=1}^{N}a_{i}x(n-i)+e(n)\]
+Linear Prediction (LPC)
+\begin_inset LatexCommand \index{linear prediction}
\end_inset
\layout Standard
-In the
-\emph on
-z
-\emph default
--domain, this can be expressed as
-\layout Standard
-
\begin_inset Formula \[
-x(z)=\frac{1}{A(z)}\: e(z)\]
+y(n)=\sum _{i=1}^{N}a_{i}x(n-i)\]
\end_inset
-\layout Standard
+\begin_inset Formula \[
+e(n)=x(n)-y(n)=x(n)-\sum _{i=1}^{N}a_{i}x(n-i)\]
-where
-\begin_inset Formula $A(z)$
\end_inset
- is defined as
-\layout Standard
+
+\begin_inset Formula \[
+E=\sum _{n=0}^{L-1}\left[e(n)\right]^{2}=\sum _{n=0}^{L-1}\left[x(n)-\sum _{i=1}^{N}a_{i}x(n-i)\right]^{2}\]
+
+\end_inset
\begin_inset Formula \[
-A(z)=1-\sum _{i=1}^{N}a_{i}z^{-i}\]
+\frac{\partial E}{\partial a_{i}}=\frac{\partial }{\partial a_{i}}\sum _{n=0}^{L-1}\left[x(n)-\sum _{i=1}^{N}a_{i}x(n-i)\right]^{2}=0\]
\end_inset
\layout Standard
-We usually refer to
-\begin_inset Formula $A(z)$
+The
+\begin_inset Formula $a_{i}$
\end_inset
- as the analysis filter and
-\begin_inset Formula $1/A(z)$
+ filter coefficients are computed using the Levinson-Durbin
+\begin_inset LatexCommand \index{Levinson-Durbin}
+
\end_inset
- as the synthesis filter.
-\layout Standard
+ algorithm, which starts from the auto-correlation
+\begin_inset LatexCommand \index{auto-correlation}
-The
-\begin_inset Formula $A(z)$
\end_inset
- filter is computed using the Levinson-Durbin algorithm, which starts from
- the auto-correlation
+
\begin_inset Formula $R(m)$
\end_inset
@@ -256,7 +292,7 @@
\begin_inset Formula \[
-r(m)=\sum _{i=0}^{N-1}x(i)x(i-m)\]
+R(m)=\sum _{i=0}^{N-1}x(i)x(i-m)\]
\end_inset
@@ -270,10 +306,10 @@
filter, we have:
\begin_inset Formula \[
\mathbf{R}=\left[\begin{array}{cccc}
- r(0) & r(1) & \cdots & r(N-1)\\
- r(1) & r(0) & \cdots & r(N-2)\\
+ R(0) & R(1) & \cdots & R(N-1)\\
+ R(1) & R(0) & \cdots & R(N-2)\\
\vdots & \vdots & \ddots & \vdots \\
- r(N-1) & r(N-2) & \cdots & r(0)\end{array}
+ R(N-1) & R(N-2) & \cdots & R(0)\end{array}
\right]\]
\end_inset
@@ -281,8 +317,8 @@
\begin_inset Formula \[
\mathbf{r}=\left[\begin{array}{c}
- r(1)\\
- r(2)\\
+ R(1)\\
+ R(2)\\
\vdots \\
R(N)\end{array}
\right]\]
@@ -327,7 +363,7 @@
This is in theory; in practice because of finite precision, there are two
commonly used techniques to make sure we have a stable filter.
First, we multiply
-\begin_inset Formula $r(0)$
+\begin_inset Formula $R(0)$
\end_inset
by a number slightly above one (such as 1.0001), which is equivalent to
@@ -336,14 +372,76 @@
to filtering in the frequency domain, reducing sharp resonances.
\layout Standard
+The linear prediction model represents each speech sample as linear combination
+ of past samples, plus an error signal called the excitation (or residual).
+\begin_inset Formula \[
+x(n)=\sum _{i=1}^{N}a_{i}x(n-i)+e(n)\]
+
+\end_inset
+
+
+\layout Standard
+
+In the
+\emph on
+z
+\emph default
+-domain, this can be expressed as
+\layout Standard
+
+
+\begin_inset Formula \[
+x(z)=\frac{1}{A(z)}\: e(z)\]
+
+\end_inset
+
+
+\layout Standard
+
+where
+\begin_inset Formula $A(z)$
+\end_inset
+
+ is defined as
+\layout Standard
+
+
+\begin_inset Formula \[
+A(z)=1-\sum _{i=1}^{N}a_{i}z^{-i}\]
+
+\end_inset
+
+
+\layout Standard
+
+We usually refer to
+\begin_inset Formula $A(z)$
+\end_inset
+
+ as the analysis filter and
+\begin_inset Formula $1/A(z)$
+\end_inset
+
+ as the synthesis filter.
+\layout Standard
+
Because LPC coefficients have very little robustness to quantization, they
- are converted to Line Spectral Pair (LSP) coefficients which have a much
- better behaviour with quantization, one of them being that it's easy to
- keep the filter stable.
+ are converted to Line Spectral Pair
+\begin_inset LatexCommand \index{line spectral pair}
+
+\end_inset
+
+ (LSP) coefficients which have a much better behaviour with quantization,
+ one of them being that it's easy to keep the filter stable.
\layout Subsection
Pitch Prediction
+\begin_inset LatexCommand \index{pitch}
+
+\end_inset
+
+
\layout Standard
During voiced segments, the speech signal is very periodic, so it is possible
@@ -405,6 +503,16 @@
\layout Subsection
Analysis-by-Synthesis and Error Weighting
+\begin_inset LatexCommand \index{error weighting}
+
+\end_inset
+
+
+\begin_inset LatexCommand \index{analysis-by-synthesis}
+
+\end_inset
+
+
\layout Standard
Most (if not all) modern audio codecs attempt to
@@ -488,6 +596,11 @@
\end_inset
+\begin_inset LatexCommand \index{narrowband}
+
+\end_inset
+
+
\layout Standard
This section looks at how Speex works for narrowband (
@@ -518,12 +631,21 @@
\layout Subsection
LPC Analysis
+\begin_inset LatexCommand \index{linear prediction}
+
+\end_inset
+
+
\layout Standard
An LPC analysis is first performed on a (Hamming) window that spans all
the current frame and half a frame in advance.
- The LPC coefficients are then converted to Line Spectral Pair (LSP), a
- representation that is more robust to quantization.
+ The LPC coefficients are then converted to Line Spectral Pair
+\begin_inset LatexCommand \index{line spectral pair}
+
+\end_inset
+
+ (LSP), a representation that is more robust to quantization.
The LSP's are considered to be associated to the
\begin_inset Formula $4^{th}$
\end_inset
@@ -569,6 +691,11 @@
\layout Subsection
Pitch Prediction (adaptive codebook)
+\begin_inset LatexCommand \index{pitch}
+
+\end_inset
+
+
\layout Standard
Speex uses a 3-tap prediction for pitch.
@@ -1674,8 +1801,12 @@
\layout Standard
-So far, no MOS (mean opinion score) subjective evaluation has been performed
- for Speex.
+So far, no MOS (Mean Opinion Score
+\begin_inset LatexCommand \index{mean opinion score}
+
+\end_inset
+
+) subjective evaluation has been performed for Speex.
In order to give an idea of the quality achivable with it, table
\begin_inset LatexCommand \ref{cap:quality_vs_bps}
@@ -1689,8 +1820,12 @@
quality sometimes varies depending on the input.
Note that the complexity is only approximate (withing 0.5 mflops and using
the lowers complexity setting).
- Decoding requires approximately 0.5 mflops in most modes (1 mflops with
- perceptual enhancement).
+ Decoding requires approximately 0.5 mflops
+\begin_inset LatexCommand \index{complexity}
+
+\end_inset
+
+ in most modes (1 mflops with perceptual enhancement).
\layout Standard
@@ -1723,7 +1858,12 @@
\layout Standard
-Bitrate (bps)
+Bit-rate
+\begin_inset LatexCommand \index{bit-rate}
+
+\end_inset
+
+ (bps)
\end_inset
</cell>
<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
@@ -1732,6 +1872,11 @@
\layout Standard
mflops
+\begin_inset LatexCommand \index{complexity}
+
+\end_inset
+
+
\end_inset
</cell>
<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
@@ -2306,6 +2451,11 @@
\layout Subsection
Perceptual enhancement
+\begin_inset LatexCommand \index{perceptual enhancement}
+
+\end_inset
+
+
\layout Standard
This part of the codec only applies to the decoder and can even be changed
@@ -2346,6 +2496,11 @@
\layout Section
\pagebreak_top
Speex wideband mode (sub-band CELP)
+\begin_inset LatexCommand \index{wideband}
+
+\end_inset
+
+
\layout Standard
For wideband, the Speex approach uses a
@@ -2360,7 +2515,12 @@
\emph on
f
\emph default
-ilter (QMF) to split the band in two.
+ilter
+\begin_inset LatexCommand \index{quadrature mirror filter}
+
+\end_inset
+
+ (QMF) to split the band in two.
The 16 kHz signal is thus divided into two 8 kHz signals, one representing
the low band (0-4 kHz), the other the high band (4-8 kHz).
The low band is encoded with the narrowband mode described in section
@@ -2895,6 +3055,11 @@
\emph on
speexenc
+\begin_inset LatexCommand \index{speexenc}
+
+\end_inset
+
+
\layout Standard
The encoder takes the following options:
@@ -2939,6 +3104,11 @@
\emph on
speexdec
+\begin_inset LatexCommand \index{speexdec}
+
+\end_inset
+
+
\layout Standard
The decoder takes the following options:
@@ -2965,7 +3135,17 @@
(-v) Print version information
\layout Section
\pagebreak_top
-Programming with Speex (the libspeex API)
+Programming with Speex (the libspeex
+\begin_inset LatexCommand \index{libspeex}
+
+\end_inset
+
+ API
+\begin_inset LatexCommand \index{API}
+
+\end_inset
+
+)
\layout Subsection
Encoding
@@ -3345,6 +3525,11 @@
\layout Subsection
Packing and in-band signalling
+\begin_inset LatexCommand \index{in-band signalling}
+
+\end_inset
+
+
\layout Standard
Sometimes it is desirable to pack more than one frame per packet (or other
@@ -3877,6 +4062,11 @@
\layout Section
\pagebreak_top
Formats and standards
+\begin_inset LatexCommand \index{standards}
+
+\end_inset
+
+
\layout Standard
Speex can encode speech in both narrowband and wideband and provides different
@@ -3920,7 +4110,12 @@
some modes.
\layout Subsection
-RTP Payload Format
+RTP
+\begin_inset LatexCommand \index{RTP}
+
+\end_inset
+
+ Payload Format
\layout Standard
This is a work in progress.
@@ -3934,7 +4129,12 @@
\layout Subsection
-Ogg file format
+Ogg
+\begin_inset LatexCommand \index{Ogg}
+
+\end_inset
+
+ file format
\layout Standard
Speex bit-streams can be stored in Ogg files.
@@ -4493,7 +4693,17 @@
FAQ
\layout Subsection*
-Vorbis is open-source and patent-free, why do we need Speex?
+Vorbis is open-source
+\begin_inset LatexCommand \index{open-source}
+
+\end_inset
+
+ and patent-free
+\begin_inset LatexCommand \index{patent}
+
+\end_inset
+
+, why do we need Speex?
\layout Standard
Vorbis is a great project but its goals are not the same as Speex.
@@ -4503,7 +4713,17 @@
typically 2-4 times higher compression at equal quality.
\layout Subsection*
-Ogg, Speex, Vorbis, what's the difference?
+Ogg
+\begin_inset LatexCommand \index{Ogg}
+
+\end_inset
+
+, Speex, Vorbis
+\begin_inset LatexCommand \index{Vorbis}
+
+\end_inset
+
+, what's the difference?
\layout Standard
Ogg is a
@@ -4531,7 +4751,12 @@
Actually, if what you do is Voice of IP (VoIP), you don't need Ogg at all.
\layout Subsection*
-Can I use Speex for compressing music?
+Can I use Speex for compressing music
+\begin_inset LatexCommand \index{music}
+
+\end_inset
+
+?
\layout Standard
Just like Vorbis is not really adapted to speech, Speex is really not adapted
@@ -4622,7 +4847,12 @@
causing clipping when saving as 16-bit PCM.
\layout Subsection*
-Can Speex pass DTMF?
+Can Speex pass DTMF
+\begin_inset LatexCommand \index{DTMF}
+
+\end_inset
+
+?
\layout Standard
I guess it all depends on the bit-rate used.
@@ -4657,7 +4887,12 @@
\layout Subsection*
-CELP, ACELP, what's the difference?
+CELP, ACELP
+\begin_inset LatexCommand \index{ACELP}
+
+\end_inset
+
+, what's the difference?
\layout Standard
CELP stands for
@@ -5113,4 +5348,12 @@
If the Document does not specify a version number of this License, you
may choose any version ever published (not as a draft) by the Free Software
Foundation.
+\layout Standard
+
+
+\begin_inset LatexCommand \printindex{}
+
+\end_inset
+
+
\the_end
<p><p>--- >8 ----
List archives: http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'cvs-request at xiph.org'
containing only the word 'unsubscribe' in the body. No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.
More information about the commits
mailing list