[vorbis-dev] joint multichannel coding (long message)

[Gabriel Bouvigne]

I've got an idea for joint multichannel coding. Here is my idea:

My idea is a process for coding 2,3,4 or 5 combined channels, in a
compatible way for decoders unable to deal with combined channels.

part1: All channels to be coded are summed, each one effected by a
coefficient related to its sound power importance for the listener. The sum
of all channels is devided by the square root of the sum of coefficients, in
order to provide an overall channel providing the user the feeling of a
monophonic coding from all channels. This overall channel is stored in a
backward compatible bitstream like if it was a monophonic recording.
Additional information to reconstruct all the joint coded channels will be
put in ancillary data. This way, old decoders are able to provide an acurate
monophonic overview of the data.

part2: this scheme can be applied to front left and front right,
or front left, front right and middle,
or front left, front right, rear left and rear right,
or front left, front right, middle, rear left and rear right
multichannel configurations. Information will be used to inform the decoder
of the multichannel configuration used. As an example, a set of 2 bits could
be used to indicate which of the 4 configurations is used.

part3: In case of 2 or 3 channels configuration, 1 additionnal signal will
be used to ensure reconstruction, and in case of 4 or 5 channels
configuration, 2 additionnal signals will be used to ensure reconstruction.

part4: For the front part, the additionnal data will include a front
difference channel, indicating the difference signal to the overall channel
necessary to reconstruct those two channels.

part5: If the multichannel configuration includes rear channels, another
additionnal signal similar to the one used in part4 will be used to
reconstruct rear channels.

part6: As such a coding scheme could, in case of strong channels separation,
lead to a decrease of coding efficiency, the encoding process can choose to
use only a low resolution representation of the overall channel and increase
the resolution of the difference channels. A little coding efficiency could
be lost in some specific parts due to this scheme, but the overall encoding
will be more efficient than without multichannel combining, and the signal
will still be backward compatible.

patents discussion:

*US5285498:
This patent covers stereo, while my idea covers up to 5 audio channels. This
patent discribes switching between L/R and M/S coding, but my idea always
uses joint encoding to ensure backward compatibility.

*US5701346
This patent uses at least 2 signals to provide signals that are compatible
with existing decoders. My idea uses only 1 signal for backward
compatibility.

*US5703999
This patent uses joint coding or basic coding based on coding efficiency. My
idea always uses joint coding, even if it could use a lot more bits for
difference channels than for overall channel.

Regards,


--

Gabriel Bouvigne - France
bouvigne at mp3-tech.org
mobile phone: gsm at mp3-tech.org
icq: 12138873

MP3' Tech: www.mp3-tech.org
personal page: gabriel.mp3-tech.org

--- >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.