Hi Bill,
Thanks, and yeah I definitely agree that would be a lot of testing! I
considered doing a simple difference of the individual sample values to
give a rough estimate at the difference, but the metric you suggest
sounds much better.
I don't have matlab, but I don't mind implementing that equation in C.
Seems like it could become useful. I'll post back to the list when I
have a program and values for those.
73 KG4SGP - Jim
On 10/26/2012 07:05 AM, Bill Cowley wrote:
Hi Jim,
A nice demo! I agree that listening to the results is going to be the
best option for measuring the bit sensitivities. But if there's lots
of bits, over different samples, perhaps tested with different channel
codes, then that's a lot of listening!
So it might be interesting to test a quantitative measure of the
effect of different bit errors and see how that compares to the
subjective test. Spectral distance measures can be used in this
situation eg see http://en.wikipedia.org/wiki/Log-spectral_distance
<http://en.wikipedia.org/wiki/Log-spectral_distance> The aim would
be to measure the average distortion, in terms of power spectral
differences, over time periods that match the variations in typical
speech. Errors could be introduced randomly and used to rank bit
sensitivities, for a given BER.
There's various code around eg
http://www.mathworks.com.au/matlabcentral/fileexchange/9998-log-spectral-distance.
I've used other spectral distance measures but not this one. Not sure
how well it would work, but could be useful for coarse tuning of the
error correction ...
Cheers, Bill
On 26 October 2012 19:14, David Rowe <[email protected]
<mailto:[email protected]>> wrote:
Hi Jimmy,
Very good work .... nice to be able to hear the effect f each bit
getting messed up.
Cheers,
David
On Fri, 2012-10-26 at 03:58 -0400, Jimmy Carter wrote:
> Hello again,
>
> I went ahead and tried to make it a little easier for others to help
> with the subjective measurements.
>
> A small write-up can be found at
http://kg4sgp.com/codec2-bit-study.html
> . This page has links to audio with corresponding bit
corruptions. From
> the listening I've done there is definitely a difference in the
> sensitivity of the different bit types.
>
> Please excuse the simple site. As you can probably tell it's not a
> priority of mine.
>
> KG4SGP - Jim
>
------------------------------------------------------------------------------
Everyone hates slow websites. So do we.
Make your web apps faster with AppDynamics
Download AppDynamics Lite for free today:
http://p.sf.net/sfu/appdyn_sfd2d_oct
_______________________________________________
Freetel-codec2 mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
WINDOWS 8 is here.
Millions of people. Your app in 30 days.
Visit The Windows 8 Center at Sourceforge for all your go to resources.
http://windows8center.sourceforge.net/
join-generation-app-and-make-money-coding-fast/
_______________________________________________
Freetel-codec2 mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
