|
|
| Pete Fraser 2006-01-20, 9:55 pm |
| For the last ten years various papers have been touting
the advantages of EREC (Error Resilient Entropy Coding).
It seems to offer a substantial advantage in image quality
with a lossy channel, requires a minimal increase in
bit-rate, and is low-complexity. However, nobody seems
to be using it in real life. Are there di vantages that
the papers don't address that would dissuade folks from using
it, or is it actually used, and I'm just not aware of it?
| |
| Michael Schöberl 2006-01-23, 3:55 am |
| > For the last ten years various papers have been touting
> the advantages of EREC (Error Resilient Entropy Coding).
The problem: this is only usefull if you want to transmit one single
type of data over your channel and all of this data could accept some loss.
Let's say we could build an internet that can transmit images in a lossy
way ... how would you control the loss? (someone is just surfing, the
other one has med images and a third one has 10 hops in between adn
another one uses vpn)
rapid switching between lossy modes? what if you mux some more packets
together for a long-distance fiber link?
but - there are some applications ... I think this is reasonable for the
last mile and afair this used in the GSM speech compression:
the fullrate speech codec has 13kbit/s but if you want to use the
timeslot for data transmission you will only get 9.6kbit/s ...
for speech compression there is some data that gets transmitted with
error protection (or even transmitted twice) and less important doesn't
bye,
Michael
| |
| Willem 2006-01-23, 3:55 am |
| Michael wrote:
)> For the last ten years various papers have been touting
)> the advantages of EREC (Error Resilient Entropy Coding).
)
) The problem: this is only usefull if you want to transmit one single
) type of data over your channel and all of this data could accept some loss.
The *channel* is lossy, not the algorithm.
SaSW, Willem
--
Disclaimer: I am in no way responsible for any of the statements
made in the above text. For all I know I might be
drugged or something..
No I'm not paranoid. You all think I'm paranoid, don't you !
#EOT
| |
| Pete Fraser 2006-01-23, 6:55 pm |
| "Michael Schöberl" <MSchoeberl@mailtonne.de> wrote in message
news:43d4a54a$1@news.fhg.de...
>
> The problem: this is only usefull if you want to transmit one single type
> of data over your channel and all of this data could accept some loss.
> Let's say we could build an internet that can transmit images in a lossy
> way ... how would you control the loss? (someone is just surfing, the
> other one has med images and a third one has 10 hops in between adn
> another one uses vpn)
> rapid switching between lossy modes? what if you mux some more packets
> together for a long-distance fiber link?
I think it's more intended for wireless links, where there is likely
to be some intermittent loss.
It appears to have the advantage of being more robust to data
loss with a relatively minor increase in data rate (transmission of
slot lengths). It looks like it would be appropriate for any block-
based video transform coding system.
| |
| Michael Schöberl 2006-01-23, 6:55 pm |
| > I think it's more intended for wireless links, where there is likely
> to be some intermittent loss.
> It appears to have the advantage of being more robust to data
> loss with a relatively minor increase in data rate (transmission of
> slot lengths). It looks like it would be appropriate for any block-
> based video transform coding system.
There is another advantage for wireless point-to-point links (like gsm
mobile phone). You can reduce the transmit power to a level where you
just have few slight errors ...
The user would not notice as the signal does not disappear ...
I don't know if this is used in broadcasting media like DVB or DAB (I
don't think so) ... the gain would be much less as you always transmit
at full power to reach as far as possible
At least at DRM (digital radio mondiale) the signal does not degrade but
the audio just drops ...
bye,
Michael
| |
| Pete Fraser 2006-01-23, 6:55 pm |
| "Michael Schöberl" <MSchoeberl@mailtonne.de> wrote in message
news:43d509b7$1@news.fhg.de...
> There is another advantage for wireless point-to-point links (like gsm
> mobile phone). You can reduce the transmit power to a level where you just
> have few slight errors ...
> The user would not notice as the signal does not disappear ...
Is it used in GSM? I see a lot of papers talking about it, but who actually
uses it? Is there a divantage that none of the papers addresses,
that has discouraged its use?
>
> I don't know if this is used in broadcasting media like DVB or DAB (I
> don't think so) ... the gain would be much less as you always transmit at
> full power to reach as far as possible
> At least at DRM (digital radio mondiale) the signal does not degrade but
> the audio just drops ...
| |
| Michael Schöberl 2006-01-24, 7:55 am |
| > Is it used in GSM? I see a lot of papers talking about it, but who actually
> uses it? Is there a divantage that none of the papers addresses,
> that has discouraged its use?
have a look at
http://ccnga.uwaterloo.ca/~jscouria/GSM/gsmreport.html
chapter Channel coding and modulation
the bits from the RPE-LPC speech codec are divided into
3 classes ... each class has a different scheme for error
detection/correction
how I understodd this: you take a 20 ms sample of your
signal ... let's say 1280 bit (64 kBit/s input signal)
those get compressed to 260 Bit (13kbit/s)
depending on the meaning of those bits some get better
protection and some get no protection at all ...
after error detection/protection you have 456 Bits that
get transmitted
it's not really the entropy coder that does the error
protection but the mechanism is coupled very closely
to the entropy coder ...
another example: in JPEG2000 you could still decode
most of your image if one block has bit errors.
But is this EREC?
sort of ... depends on the definition
bye,
Michael
| |
| Pete Fraser 2006-01-24, 9:55 pm |
| "Michael Schöberl" <MSchoeberl@mailtonne.de> wrote in message
news:43d5faff$1@news.fhg.de...
> another example: in JPEG2000 you could still decode
> most of your image if one block has bit errors.
>
> But is this EREC?
> sort of ... depends on the definition
It's not really EREC, if I use the definition in the 1996
Redmill and Kingsbury paper.
|
|
|
|