Is there a standard set of test vectors for an MP2 decoder, and are they expected to give bit exact results? Steve On 01/04/2015 01:27 AM, Thomas Auge wrote:
A very interesting tech demo, but from his own readme, it might not be suitable for a freeswitch module:
- imprecise: due to the integer-only implementation, many rounding errors will occur; expect a 45-55 dB noise margin! - slow: it implements the synthesis process from the spec, which is very slow (82 multiplications, 1-2 divisions and countless shifts and additions per sample!) - not robust: it may easily crash on broken input data. Also, it ignores CRC values completely. - not thread-safe
A bit of a bummer, because a lightweight decoder on top of twolame would have been nice.
On 03.01.2015 13:51, Steve Underwood wrote:
I have no idea how well it works, but there is this MP2 decoder - http://keyj.emphy.de/kjmp2/
On 01/03/2015 07:00 AM, Brian West wrote:
That looks like just the encoder, so we would need both the encoder and decoder that takes signed linear. I'll be on the look out for an option too.
On Fri, Jan 2, 2015 at 4:54 PM, Thomas Auge <auge@virtues.net <mailto:auge@virtues.net>> wrote:
> G722 can be a native bearer capability on an ISDN b-channel but thats not widely used. MP2 in your case is just > establishing a data channel be it via ISDN or IP.
Yup. I'm fairly certain that's what FreeTDM does for G722, else it wouldn't make much sense to have a codec configuration option. :)
> Before proceeding I would seek out a license compatible encoder/decoder library that can be used, MP2 can mean so many > things their are multiple levels, options and details required to properly implement it. I highly suspect its not an > afternoon event, I could be wrong if the right library is found thats license compatible.
There are a couple, T(w/o)oLame comes to mind. (http://www.twolame.org/ LGPL licensed) Any pointers as to which capabilities are needed in such a library? It seems the libraries are as outdated as the codec. :)