[vorbis] Beta3 impressions

[Gregory Maxwell]

On Wed, Nov 29, 2000 at 04:47:38PM -0600, brian wrote:
[snip] 
>    (2) Instead of transforming time-signal data, look at the frequency
>        information spit out of the fft or preferable method.  Perform some
>        sort of wavelet packet analysis and reconstruct an approximation of
>        the frequency signal using the psychoacoustic model to direct
>        precision.  Granted, this would simply be quantization via
>        wavelet packets.  However, it would be interesting to see what
>        happens when performed.  (It would obviously not solve the
>        problem of pre-echo control, but may be an interesting thing
>        to investigate altogether).
> 
> I was curious to the methods that Monty had in mind (as far as choosing
> bases, etc).  Comments?

I did a fair amount of screwing around in this configuration before Monty
had codebooks implemented.

Basically I took the DWT of the residue (I believe spline had the most
energy in the fewest coefficients) and quantized it.

What I found was that the slight offsets and rippling it produced in the
residue when using lots of quantization produced very unpleasant results. 
Perhaps smarter quantization would have been more effective..

Really, I was only playing with it to get a handle on the code base and I
actually wanted to hear acceptable output from a file smaller then the
original. :)

Later, after codebooks were in place and Monty added sufficient hooks for
time coding, I played more, but I found that with the codec block sizes
currently in use, Monty's level normalizing took care of preecho enough that
I was unable to produce better results otherwise.

I have a suspicion the untrained entropy books might perform worse preecho
wise then trained books, but I have not tested that theory. Furthermore, I
have not spent enough time with the code since lattice books were added to
really trust my intuition on such matters.
--- >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-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.