Illustrating the bit-stream
authorJean-Marc Valin <jean-marc.valin@octasic.com>
Mon, 8 Jun 2009 21:45:00 +0000 (17:45 -0400)
committerJean-Marc Valin <jean-marc.valin@octasic.com>
Mon, 8 Jun 2009 21:45:00 +0000 (17:45 -0400)
doc/ietf/draft-valin-celt-codec.xml

index 4b8fc5a..b891b16 100644 (file)
 </address>
 </author>
 
+<author initials="T" surname="Terriberry" fullname="Timothy B. Terriberry">
+<organization>Xiph.Org Foundation</organization>
+<address>
+<postal>
+<street></street>
+<city></city>
+<region></region>
+<code></code>
+<country></country>
+</postal>
+<email>tterribe@xiph.org</email>
+</address>
+</author>
+
+<author initials="G" surname="Maxwell" fullname="Gregory Maxwell">
+<organization>Juniper Networks</organization>
+<address>
+<postal>
+<street>2251 Corporate Park Drive, Suite 100</street>
+<city>Herndon</city>
+<region>VA</region>
+<code>20171-1817</code>
+<country>USA</country>
+</postal>
+<email>gmaxwell@juniper.net</email>
+</address>
+</author>
+
 <!-- <author initials="et" surname="al." fullname="et al.">
 <organization></organization>
 </author>
@@ -144,6 +172,21 @@ alpha_p=0.8. The inverse of the pre-emphasis is applied at the decoder.</t>
 (<xref target="celt.c">celt.c</xref>).
 </t>
 
+<t>
+<figure>
+<artwork>
++-----------------+---------------------+------------------------------+
+|  Feature flags  | (pitch period if P) | (transient scalefactor if S) |
++-----------------+---------------------+------------------------------+
+|  (transient time if scalefactor == 3) |  coarse energy               |
++----------------+----------------------+------------------------------+
+|  fine energy   |  PVQ indices  for all bands |  end of frame pattern |
++----------------+-----------------------------------------------------+
+</artwork>
+</figure>
+</t>
+
+
 <section anchor="Range Coder" title="Range Coder">
 <t>
 derf?
@@ -209,6 +252,7 @@ In the case where the scalefactor is 1 or 2 and the mode is defined to use more
 
 <section anchor="folding" title="Spectral folding (F)">
 <t>
+The last encoding feature in CELT is spectral folding. It is designed to prevent "birdie" artefacts caused by the sparse spectra often generated by low-bitrate transform codecs.
 </t>
 </section>
 
@@ -222,7 +266,6 @@ frequency-domain samples. A "low-overlap" window is used to reduce the algorithm
 It is derived from a basic (with full overlap) window that is the same as the one used in the Vorbis codec: W(n)=[sin(pi/2*sin(pi/2*(n+.5)/L))]^2. The low-overlap window is created by zero padding the basic window and inserting ones in the middle, such that the resulting window still satisfies power complementarity. The MDCT is computed in mdct_forward() 
 (<xref target="mdct.c">mdct.c</xref>), which includes the windowing operation.
 </t>
-
 </section>
 
 <section anchor="Bands and Normalization" title="Bands and Normalization">