f

#### To Code or Not to Code?

Just read a paper with the very interesting title, "To code or not to
code: lossy source-channel communication revisited", Gastpar et al,
IEEE Trans Information Theory, May 2003, p 1147.

http://www.eecs.berkeley.edu/~gastpar/01197846.pdf

As far as I can tell, the idea is that although source and channel
coding can be separated without loss of optimality (Shannon) it can be
overly complex, and less complex systems can get close to this
opimality by combining source and channel coding, or at least matching
them in some way.

This sounds like a pretty interesting idea, unfortunately my
information theory knowhow lies way below that of my signal processing.
More precisely, I cant understand a word of the paper! I dont suppose
any budding info theorists could shed a bit of light for me?

Slainte
Porterboy

 0
porterboy76 (215)
7/8/2005 7:30:16 AM
comp.dsp 20333 articles. 1 followers. allnor (8510) is leader.

6 Replies
2204 Views

Similar Articles

[PageSpeed] 34

porterboy76@yahoo.com wrote:
> Just read a paper with the very interesting title, "To code or not to
> code: lossy source-channel communication revisited", Gastpar et al,
> IEEE Trans Information Theory, May 2003, p 1147.
>
> http://www.eecs.berkeley.edu/~gastpar/01197846.pdf
>
> As far as I can tell, the idea is that although source and channel
> coding can be separated without loss of optimality (Shannon) it can be
> overly complex, and less complex systems can get close to this
> opimality by combining source and channel coding, or at least matching
> them in some way.

Hmmm... being pathologically interested in practical methods of
implementations, I tend to be a bit cautious about signal
processing schemes where the transmission channels is assumed to
be know or measurable, and then corrected for.

To my very naive mind, "channel coding" is based on the assumption
that there is some sort of independent method of communications
between source and reciever in time-variable channels (how else
can one know exactly what was sent and what ought to be recieved?)
or the channel must be assumed to be static. In the former case,
why not use this other channel for communication in the first place?
In the latter case, why does one need channel coding, why not use
source coding alone?

Again, I am a very naive person and like you, I don't have a clue
what I am talking about what information theory is concerned.

Rune

 0
allnor (8510)
7/8/2005 1:45:29 PM
<porterboy76@yahoo.com> wrote in message
> This sounds like a pretty interesting idea, unfortunately my
> information theory knowhow lies way below that of my signal processing.
> More precisely, I cant understand a word of the paper! I dont suppose
> any budding info theorists could shed a bit of light for me?

I only read the introduction (It looks interesting, so I'll read the rest
later), but I think I can offer an enlightening example.

Let's say you are transmitting real-time video.  The normal approach is to
choose a method of channel encoding that lets you transmit reliably at rates
close to the channel capacity.  Then you use a lossy compression scheme to
compress the video down to that rate by discarding less important
information.  If both these jobs are done well, then this approach achieves
an optimal distortion level in the video for the transmission rate, but it's
complicated and the channel coding can require long codes, which means long
delays and lots of processing.

The proposed alternative is to combine the to coding stages (compression and
information so that you can fit the signal into the channel capacity, you
transmit ALL of the information, but allocate signal power and reliability
to the various pieces of information according to how important they are, so
that the channel itself will transmit the most important information
relatively reliably and let the less important stuff wash out with noise.

The second approach can be a lot easier than the two-stage approach above,
and can involve much less delay, because you don't have to go through the
effort of making the channel very reliable.  The author seems to be
suggesting that you can achieve the optimal distortion level with the second
approach as well as you can with the first.

--
Matt

 0
mt0000 (280)
7/9/2005 2:16:17 AM
> information so that you can fit the signal into the channel capacity, you
> transmit ALL of the information, but allocate signal power and reliability
> to the various pieces of information according to how important they are, so
> that the channel itself will transmit the most important information
> relatively reliably and let the less important stuff wash out with noise.

Well, that seems like a smart idea alright. But if the coding is
dependent on the source information, then would the delay and
complexity introduced in coding for every new source of information not
make this approach MORE complex rather than less complex? In the
traditional approach, the source coding increases signal entropy, so in
effect we transmit a white signal, while the channel coding accounts
for the worst expected channel noise. These seem like well defined and
not so complex problems, compared with trying to work out what part of
a sources information is least (most) likely to be hit in the channel,
and applying a weaker (stronger) source/channel code to that part of
the information.

Also, I gather from the conclusion, that the authors approach is not
constructive: it says less complex source/channel coding is possible,
but does not say how it can be done (I think! Could be wrong there...).

 0
porterboy76 (215)
7/11/2005 7:25:21 AM
Hm.

Is this not, in a way, already used in ordinary modems? With a
theoretical limit of 64kbps, they get up to 57kbps, which is quite
close. And this is lossless, right?
I have no idea how they manage to get this close with relatively poor
line qualities...

Andre

porterboy76@yahoo.com wrote:

>>information so that you can fit the signal into the channel capacity, you
>>transmit ALL of the information, but allocate signal power and reliability
>>to the various pieces of information according to how important they are, so
>>that the channel itself will transmit the most important information
>>relatively reliably and let the less important stuff wash out with noise.
>
>
> Well, that seems like a smart idea alright. But if the coding is
> dependent on the source information, then would the delay and
> complexity introduced in coding for every new source of information not
> make this approach MORE complex rather than less complex? In the
> traditional approach, the source coding increases signal entropy, so in
> effect we transmit a white signal, while the channel coding accounts
> for the worst expected channel noise. These seem like well defined and
> not so complex problems, compared with trying to work out what part of
> a sources information is least (most) likely to be hit in the channel,
> and applying a weaker (stronger) source/channel code to that part of
> the information.
>
> Also, I gather from the conclusion, that the authors approach is not
> constructive: it says less complex source/channel coding is possible,
> but does not say how it can be done (I think! Could be wrong there...).
>

--

 0
7/11/2005 3:55:23 PM
<porterboy76@yahoo.com> wrote in message
> Well, that seems like a smart idea alright. But if the coding is
> dependent on the source information, then would the delay and
> complexity introduced in coding for every new source of information not
> make this approach MORE complex rather than less complex?

Yes, it's not appropriate for general use digital channels, but seems to
make a lot of sense for channels that are dedicated to a single purpose that
involves lossy compression in normal 2-stage designs.  HDTV, digital radio,
cellular voice communication, etc., are all fine examples.

> Also, I gather from the conclusion, that the authors approach is not
> constructive: it says less complex source/channel coding is possible,
> but does not say how it can be done (I think! Could be wrong there...).

It would be different for every medium.

--
Matt

 0
mt0000 (280)
7/12/2005 1:13:07 AM
> Is this not, in a way, already used in ordinary modems? With a
> theoretical limit of 64kbps, they get up to 57kbps, which is quite
> close. And this is lossless, right?
> I have no idea how they manage to get this close with relatively poor
> line qualities...

(delayed reply... was out for a week...)

You are talking about voice band modems, which have a relatively flat
frequency response, and good SNR. Also, because of the low information
rates, there is time to do a lot of signal processing such as
constellation shaping, turbo coded modulation with iterative decoding,
and precise decision feedback equalisation. However, all of these
things are generally performed independently of the source coding,
which I believe is undertaken at a higher level in the OSI model. I
think the point of the paper in question is that source and channel
coding can be combined. Anyway, I'm not so sure that the scheme has
practical value...

 0
porterboy76 (215)
7/21/2005 7:43:28 AM

Similar Artilces:

To code or not to code?
To code or not to code, that is question! Reporting from Lotus Notes Applications has never been easy. Notes developers usually need to write reams of code to create even the simplest of reports. Check this - a single-level cross-tab report takes about 2 days of code writing! Creating reports with higher functionality only means more coding and more complexities. Is there an alternative to this? You bet there is! Introducing IntelliPRINT Reporting - a cutting-edge next generation solution for Lotus Notes reporting. It is a native Lotus Notes solution that acts as an add-on lay...

To code or not to code?
To code or not to code, that is question! Reporting from Lotus Notes Applications has never been easy. Notes developers usually need to write reams of code to create even the simplest of reports. Check this - a single-level cross-tab report takes about 2 days of code writing! Creating reports with higher functionality only means more coding and more complexities. Is there an alternative to this? You bet there is! Introducing IntelliPRINT Reporting - a cutting-edge next generation solution for Lotus Notes reporting. It is a native Lotus Notes solution that acts as an add-on lay...

good code, bad code, magic code
i've discovered (on my server) that it doesn't seem to be necessary to call data from a form via $_POST, or to retrieve cookies via$_COOKIE, and other such things; they are available by name alone. however, i've heard that just because you CAN do something, doesn't mean you SHOULD. my question: is this considered bad practice? also, when submitting a file via <input name="file" type="file" /> i find that not only does $file get set, but$file_name and others also "magically" get set... I guess I just find that odd coming from C++ where nothing seems to do work for you... quite nice and handy though :) >i've discovered (on my server) that it doesn't seem to be necessary to >call data from a form via $_POST, or to retrieve cookies via$_COOKIE, >and other such things; they are available by name alone. however, i've register_globals is a security problem. Don't write code that depends on it. >heard that just because you CAN do something, doesn't mean you SHOULD. >my question: is this considered bad practice? Yes. And you'll probably find some hosting companies that WILL NOT turn on register_globals for any price. Gordon L. Burditt great! thanks. that's all I wanted to know. ...

Code to display my code
Hi all, I am taking a class in PHP and thought that it would be nice to display the PHP code that generated my pages when the user clicked a link (or submit button). I have seen page that will display the underlying HTML on the page and it would be a nice touch...I think Is there a way to do this? TIA, Miki Michelle wrote: > I am taking a class in PHP and thought that it would be nice to display > the PHP code that generated my pages when the user clicked a link (or > submit button). I have seen page that will display the underlying HTML > on the page and it would be a nice touch...I think > > Is there a way to do this? http://www.php.net/highlight_file Example: <?php echo '<p>Hello World!</p>'; highlight_file($_SERVER['PHP_SELF']); ?> The HTML output is (reformatted): <p>Hello World!</p><code> <font color="#000000"> <font color="#0000CC">&lt;?php<br /></font> <font color="#006600">echo </font> <font color="#CC0000">'&lt;p&gt;Hello World!&lt;/p&gt;'</font> <font color="#006600">];<br/></font> <font color="#0000CC">highlight_file</font> <font color="#006600">(</font> <font color="#0000CC">$_SERVER</font> <font color="#006600">[</font>...

changing code by code
hi y'all! i want to implement a few macros to help me do my coding; when calling a macro: how i can i find out in which module i'm currently in; at which position and what text i have selected. it's gotta be possible somehow, but the screen-objekt really isn't of much help. can anyone help me? thanks in advance! mike maercker I don't know about macros. You might be able to use them to run code you write. But if you are up to coding, you need to set a reference to the 'Microsoft Visual Basic for Applications Extensibility 5.3'. It provides the interface for modifying/returning information about code modules. The CodePane object returns information about the code/module being viewed in the IDE. CodePane.CodeModule property will tell you which module is being viewed in the in the IDE. Using the CodePane's GetSelection method you can retrieve information about the currently selected text. The CodeModule object allows to locate, identify, modify lines of code in a particular module. I have not checked to see what kind of documentation Microsoft has on programming using the IDE object. You can use the object browser to look at it and retrieve information about its object model, properties and methods. Some Wrox Publishing books contain reference information about the IDE object. The "Excel 2000 VBA Programmer's Reference" does and it is what 1st introduced me the IDE object. I don't have my copy handy ( it is packed ...

Code to compile a code
In a c program I have a var with the name of a file that I want to compile, how would I compile it using c code? Profetas <xuxu_18@yahoo.com> wrote: > In a c program I have a var with the name of a file that I want to compile, > how would I compile it using c code? Probably by cobbling together a string with the command to invoke the compiler, using your variable, and than feeding it to the system() function. Regards, Jens -- \ Jens Thoms Toerring ___ Jens.Toerring@physik.fu-berlin.de \__________________________ http://www.toerring....

Conversion from a MATLAB code to a C code used by a DSP processor
Hello, I need to implement image processing in a Digital Signal Processor. I have a code ready in MATLAB, I have tried using MATLAB Coder, but its unable to go past the first line: the imread() function. I have installed the C compiler from MATLAB and that is all. How do I go about the whole of the conversion process? Do I need to install something else? ...

Automatic "code-pattern" recognition, in DSP codes
Hi I am researching in compilers and source to source code transformations, for Digital Signal Processors, in particular with the Texas Instruments Code Composer Studio 2.2 for the 55x family. I was wondering if somebody has ever heard about something like "automatic typical DSP-like kernel recognition", for example, as the compiler reads the program and sets up the data and control graphs, recognise that for(i=0, acc = 0; i < N; i++) { acc = h[n] * X[N-n]; } can be a FIR filter, that may be substituted as a whole by a fir() library function call, most optimised. Can be also understood as pattern-search in the code control/data graphs? So far I have been doing search for the subject, but found nothing, so any help, comment, or pointed direction will be much appreciated. -- ------------------------------------- Miguel Cas�s-S�nchez Room C0.07, Physics Building Department of Computer Science, University College Dublin, Belfield, Dublin 4, Ireland. Tel: +353 1 716 2513 Fax: +353 1 269 7262 Email: miguel.casassanchez_at_ucd.ie Web page: gauss.depeca.uah.es/~mcasas =?ISO-8859-15?Q?Miguel_Cas=E1s-S=E1nchez?= <Miguel.CasasSanchez@ucd.ie> wrote: >I am researching in compilers and source to source code transformations, >for Digital Signal Processors, in particular with the Texas Instruments >Code Composer Studio 2.2 for the 55x family. > >I was wondering if somebody has ever heard about something like >"automatic typical DSP-lik...

how to use assembly code inside the C code for tigersharc dsp
hi friends, can someone pls tell me how to use the assembly code for tigersharc dsp i between C code? This message was sent using the Comp.DSP web interface o www.DSPRelated.com mmonica wrote: > hi friends, > can someone pls tell me how to use the assembly code for tigersharc dsp in > between C code? > > Every tool does this differently. Read your manual. You will probably want to have a good idea of what assembly the compiler generates for a given chunk of C. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Tim Wescott wrote: > mmonica wrot...

M code into programming C code or MATLAB programming code
hallo firends, Can any body help me write this below code into MATLAB object oriented or C code form please? M=50; % total number of periods. (choose for 1s duration) f0 = 50; % fundamental AC frequency T0 = 1/f0; %fundamental AC period T = M*T0; %Time for M periods (integer M) N=30; % sample points per period. dt = T0/N; % Sample at N points per period (integer N > 20) t = dt*[1:M*N-1]; % Sampling time array A0=120/2; % amplitude of AC signal. Divide by 2 or maybe sqrt (2) ? AC=A0*sin(f0*t*2*pi); % create AC signal. DC0=50; % DC amplitude DC=DC0*ones(1,length...

Virtual Key Codes, Scan Codes and ASCII Codes in C
Not sure if this is the right board but it is in c... I've got a console program written in c which receives key presses and gets a Windows Virtual key code. I am trying to convert it to the exact ascii value the user has typed. so if shift is down it should be a capital otherwise lower case etc. I started with this: void printKey(int key) { printf("Key: %c\n", key); } and I got all upper case (not really suprising) so then after a bit of research and grafting I came up with this: int vk2ascii(unsigned int vk, int *s) { int scan; unsigned char state[256]; HKL layout=GetK...

Help for implementing FEC codes like Tornado codes & Online codes

SPSS code to SAS code
I have some SPSS code that I need to convert to SAS code. Since I haven't used SPSS for about 25 years, I'm less than confident that I am properly interpreting the SPSS code that I have. Is there any kind of program that translates code between the two languages, no matter how imperfectly? At least that would give me a starting point for the correct SAS code. -- Charlie Axilbund caxilbund@iasinfo.com ...

matlab code for turbo code
please someone send me matlab code for turbo code...............on my email id wizynt.94@gmail.com ...

creator codes and type codes
i just registered a couple of creator codes with apple. but their website says they don't keep track of type codes anymore. why not? isn't the whole point of that exercise to prevent conflicts? i'm pretty sure it would be unwise for me to try to claim, say, TEXT, but other than steering clear of the well-known ones, how am i to avoid trouble? i guess we'll all be moving to UTIs eventually, but i think it's a little early for that right now. In article <PM0003FB8D691809AF@remy.nashville.comcast.net>, Jøhnny Fävòrítê (it means "Wanton Sezhuan") <brunsona@newsguy.takethisout.com> wrote: > i just registered a couple of creator codes with apple. but their website > says they don't keep track of type codes anymore. why not? isn't the whole > point of that exercise to prevent conflicts? i'm pretty sure it would be > unwise for me to try to claim, say, TEXT, but other than steering clear of the > well-known ones, how am i to avoid trouble? I think the recommendation is that you use extensions, and a quick test shows that a document with both a known extension and a type code is recognized by extension. So if you provide an extension, having a type code won't change much. -Eric -- Eric Albert ejalbert@cs.stanford.edu http://outofcheese.org/ In article <PM0003FB8D691809AF@remy.nashville.comcast.net>, Jøhnny Fäv...

Trellis code in matlab code
Hi All Do you have some matlab code for the STTC which will give me some Idea. Any helping material or URL regarding the same will kindly be sent to me. I shall be very thankful to you for this act. Regards Kelecha ...

Bug in embeddable code or in my code?
Hi, When I run the code below, I manage to execute certain python code but not some simple imports. For example, the code below... (I removed some error checking/reference checks for brevity) // ... Py_SetProgramName(L"AAAAA"); Py_SetPath(L"path\\to\\python35.zip"); // embeddable from the python site. Py_Initialize(); PyThreadState *mainstate = PyThreadState_Get(); PyInterpreterState* mainInterpreterState = mainstate->interp; PyObject *main_module = PyImport_AddModule("__main__"); PyObject *main_dict = PyModule_GetDict(main_module); PyObject *local_dic = PyDict_New(); const char* s = "import ctypes\n"; PyObject * PyRes = PyRun_String(s, Py_file_input, main_dict, local_dic); PyObject* ex = PyErr_Occurred(); if (NULL != ex) { // didn't work, import error. } // cleanup... // ... I can do some imports, ("import sys" for example), but most of the others do not work, "import ctypes" for example. Any suggestions as to what I might be missing? S. ...

From wide coded to long coded?
As a student not having much experience with SPSS, I'm having trouble transforming wide-coded data to long-coded data. My dataset consists of 4900 households responding to the question whether the sick person in the illness episode was male or female. This question is repeated for 10 illness episodes. I therefore have the gender-info in wide code, but since I need the gender info for a regression analysis, I need the gender-info in long code. So instead of 4900 Household ID's, I need 49.000 person ID's with the gender of each person. However, I do not have a clue how to do this. Is there anyone that can give me a tip? Kind regards, Vanessa Vanessa, I don't know if you dealing with different illnesses, but something like this code should work for you: GET FILE=origdata. SAVE OUTFILE=ill1/KEEP=houseid,illness1,sex1/RENAME=(illness1,sex1 = illness,sex). GET FILE=origdata. SAVE OUTFILE=ill2/KEEP=houseid,illness2,sex2/RENAME=(illness2,sex2 = illness,sex). GET FILE=origdata. SAVE OUTFILE=ill3/KEEP=houseid,illness3,sex3/rRENAME=(illness3,sex3 = illness,sex). (continue GET/SAVE for all 10 illnesses/illness episodes) ADD FILES FILE=ill1/FILE=ill2/fFILE=ill3/.....FILE=ill10. At this point, you will have 3 vars - HOUSEID, ILLNESS (which could be an illness code or an illness episode #), and SEX. Hope this helps!! Sue Vanessa wrote: > As a student not having much experience with SPSS, I'm having trouble &g...

SPSS code to SAS code
Hi guys Is there a way to convert SPSS codes to SAS syntax?? I am not asking about SPSS datasets to SAS, but for codes. Thank you and looking forward to your help. Marios ...

less code best code

Coding Gain: Huffman Code
Hey All, Could someone tell me how to find out the coding gain if Huffman coding i used. The BER without coding is 0.1. I found the Huffman code of th alphabet and got an efficiency of 98 %. How do I calculate the coding gai or the BER when the coding is used for a specific SNR value? Thanks, Nisha On Sun, 09 Dec 2007 08:03:22 -0600, "Nisha_Malh" <farah727rash@gmail.com> wrote: >Hey All, > >Could someone tell me how to find out the coding gain if Huffman coding is >used. The BER without coding is 0.1. I found the Huffman code of the >alphabet and got an efficiency of 98 %. How do I calculate the coding gain >or the BER when the coding is used for a specific SNR value? > >Thanks, >Nisha > Usually 'coding gain' is expressed in dB at a particular BER. For example if a particular coded system achieves Pe = 10e-3 at 6dB SNR, and the same BER at 10dB SNR when the same system is uncoded, then the 'coding gain' under those conditions is 4dB. So typically one picks a performance metric, whether it's a particular BER or PER or something similar, and then finds the difference in SNR or Eb/No (or similar metric) for the coded and uncoded cases at the same performance level in the same channel conditions. Eric Jacobsen Minister of Algorithms Abineau Communications http://www.ericjacobsen.org >Usually 'coding gain' is expressed in dB at a particular BER. For >example if a particular coded syste...

Code Review
A geek and linux hacker from china wrote this piece of code in his / her unfinished book "the art of C programming". Under which version of C standard will it be correct? The author may think it is correct under at least some older C versions. This person submitted patches to linux kernel. Linus ever gave comment on mailing list like this to those patch code: You code is shit. Is this code really shit? /* check endianess, http://wangcong.org/down/chapter2-beta.pdf */ #include <stdio.h> #include <stdlib.h> int main(X) {exit(0 > puts((*(char *)&X == (char)X)?"Little endian":"Big endian"));} On Mon, 27 Jul 2009 19:32:33 -0700 (PDT), "lovecreatesbeauty@gmail.c0m" <lovecreatesbeauty@gmail.com> wrote: >A geek and linux hacker from china wrote this piece of code in his / >her unfinished book "the art of C programming". Under which version of >C standard will it be correct? The author may think it is correct >under at least some older C versions. > >This person submitted patches to linux kernel. Linus ever gave comment >on mailing list like this to those patch code: You code is shit. > >Is this code really shit? > >/* check endianess, http://wangcong.org/down/chapter2-beta.pdf */ > >#include <stdio.h> >#include <stdlib.h> >int main(X) C99 did away with implicit int so it is not valid under that standard. >{exit(0 > puts((*(char *)&X == ...

matlab code to protege code
hi friends i doing my project in matlab (semantic content extraction in video using fuzzy ontology) so i need help for 1) matlab code for low and high feature extraction of an image 2)how i convert matlab files(.m files) to input file for protoge software plz help me............. thanks in advance ...

matlab codes for rcpc coding
i am doing project on 60GHz application. and i want matlab codes to generate coding-decoding for rcpc channel coding... pls provide codes for rcpc.. tahnks ...

Web resources about - To Code or Not to Code? - comp.dsp

Code - Wikipedia, the free encyclopedia
"Decoded" redirects here. For the television show, see Brad Meltzer's Decoded . For the Jay-Z memoir, see Decoded (book) . A code is a rule for ...

Herding Code
While at NDC 2012 in Oslo, Jon and K. Scott talked to Shay Friedman about Roslyn, IronRuby, and the DLR. Download / Listen: 146 – Shay Friedman ...

iPhone, iPad and Mac OS X Cocoa Usability and Development Services - Instinctive Code
iPad, iPhone / iPod Touch and Mac OS X Cocoa development services and UI design.

Re/code
No Bible Available? NY Swears in Politician With iPad Bible App. By Jordan Kahn, Writer, 9to5Mac

AFL illicit drugs code: Push towards a more stringent code verges on immoral
The AFL Players Association might well be asking itself how it agreed to an Illicit drugs code so intrusive and open to capricious change in ...

Report: Text hidden deep in Facebook Messenger code hints at monetization plan
According to a report out of The Information today, Facebook has begun preparing several new features for Facebook Messenger, including a new ...

Deal: Mpow Bluetooth Headphones $26.99 W/Code, 03/24/16 If you love music and stream it from your mobile device, but you also enjoy being active and don't like the be tied down by the cords and wires ... Code Fellows moves out of South Lake Union, snags space in Belltown Dave Parker looked all over Seattle for a new space for the fast-growing coding school Code Fellows. The 20 person company – which also needs ... Dell May Announce Unit Sale to Japan's NTT Monday, Re/code Says ... announce the sale of its technology services unit to Japan’s Nippon Telegraph & Telephone Corp. for$3.5 billion as early as Monday, Re/code ...

Resources last updated: 3/30/2016 3:38:30 AM