[vorbis-dev] Beta4 LSP training

[Arnar Mar Hrafnkelsson]

Is it possible to get the commandline parameters used for training the
floor data for Beta4?  I've got the training data by defining TRAIN_LSP
and recompiling the trick is what paramters I should pass to lspvqtrain

$ ./lspvqtrain 

OggVorbis LSPdata VQ codebook trainer

<foo>vqtrain vqfile [options] [datasetfile] [datasetfile]
options: -p[arams]     <entries,dim,quant>
         -s[ubvector]  <start[,num]>
         -e[rror]      <desired_error>
         -i[terations] <maxiterations>
         -d[istance]   quantization mesh spacing for density limitation
         -b <dummy>    eliminate cell size biasing; use normal LBG

         -c <dummy>    Use centroid (not median) midpoints
examples:
   train a new codebook to 1% tolerance on datafile 'foo':
      xxxvqtrain book -p 256,6,8 -e .01 foo
      (produces a trained set in book-0.vqi)

   continue training 'book-0.vqi' (produces book-1.vqi):
      xxxvqtrain book-0.vqi

   add subvector from element 1 to <dimension> from files
      data*.m to the training in progress, prodicing book-1.vqi:
      xxxvqtrain book-0.vqi -s 1,1 data*.m

In the basic example given it's simple to use, but I want to be able to
come close to reproducing the books as they are in beta4.  I have no idea
if those additional steps have been performed. Also what does subvector
do?

-- Arnar

--- >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 'vorbis-dev-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.