[Caml-list] Signal analysis

[Caml-list] Signal analysis

[Alessandro Baretta]

Hello.

I need some advice on a project I will have to start in a 
few days. I need to write code to perform spectral analysis 
of digital signals. I do not know the details yet, but I 
know that I will have to perform Fourier trasformations on 
fairly large arrays, let us say in the neighborhood 2^16 
samples. Since the data comes from text files, I would like 
to use Ocaml/Ocamllex/Ocamlyacc, but will I be able to 
perform the numerical part at a decent speed? I realize 
"decent" here could mean just about anything, but since I 
have no real time specs, I simply have to crunch the data 
fast enough (on a fast enough PC) so as to prevent the user 
from getting exceedingly bored and frustrated between 
commands. Are there any optimized FFT (and signal analysis) 
libraries for Ocaml, or should I code the whole thing from 
scratch?

I would appreciate any ideas from anyone having some 
previous signal analysis experience with Ocaml.

Thank you very much.

Alex

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] Signal analysis

[Eray Ozkural]

Investigate the binding for FFTW library for ocaml.

fastest in the west :)

On Monday 03 June 2002 22:19, Alessandro Baretta wrote:
> Hello.
>
> I need some advice on a project I will have to start in a
> few days. I need to write code to perform spectral analysis
> of digital signals. I do not know the details yet, but I
> know that I will have to perform Fourier trasformations on
> fairly large arrays, let us say in the neighborhood 2^16

-- 
Eray Ozkural (exa) <erayo@cs.bilkent.edu.tr>
Comp. Sci. Dept., Bilkent University, Ankara
www: http://www.cs.bilkent.edu.tr/~erayo  Malfunction: http://mp3.com/ariza
GPG public key fingerprint: 360C 852F 88B0 A745 F31B  EA0F 7C07 AE16 874D 539C

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] Signal analysis

[Christophe TROESTLER]

On Mon, 3 Jun 2002, Eray Ozkural <erayo@cs.bilkent.edu.tr> wrote:
> 
> Investigate the binding for FFTW library for ocaml.

You can find it at "http://www.umh.ac.be/math/an/software.php".  The
interface is basically finished but it hasn't been thoroughly tested.
Most of the FFTW options are available -- including transforming
several arrays at once.  Do not hesitate to contact me in case of
problems or if you need further help.

Regards,
Christophe
-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] Signal analysis

[Alessandro Baretta]

Christophe TROESTLER wrote:
> On Thu, 11 Jul 2002, Alessandro Baretta <alex@baretta.com> wrote:
> 
>>Now, here I am building fft libraries and trying to get a grasp of
>>them. I have just typed "make" on my terminal, and I died horribly
>>because of a header it could not find.
> 
> 
> Do you mean the (r)fftw.h headers ?
> 

Yes. That's it. I had to specify a -I option in the Makefile 
to allow gcc to find the headers.

>>ocamlc tells me it type Complex.t is undefined. Now this seems to be
>>a real impediment: I have grep'ed my ocaml distro and have found no
>>reference to module Complex.
> 
> 
> Sorry for your inconvenience.  It's going to be said more clearly that
> the FFTW interface requires complex number hence the OCaml CVS
> distribution (which explains the strange paths in my Makefile) --
> until the next Ocaml release at least.  Don't worry, compiling Ocaml
> from CVS is easy, just look at http://camlcvs.inria.fr/ to download
> the files and follow the instructions. 

Compiling the CVS version is no real problem, but it tends 
to be slightly less stable than the distro, so I tend to 
prefer to stick with the distro for production work. Well, I 
guess I'll have to update my source directory and bootstrap it.

> I recommend running
> ./configure with an option like -prefix /usr/local/stow/ocaml-cvs/ if
> you want to keep it alongside Ocaml 3.04 (if you do not know "stow",
> have a look to it, it is interesting).

Reading your Makefile I had thought that stow was some 
strange name or acronym for a project you might be working 
on. Now you teach it is a package manager. Ok, I'll try it out.

> The compile directory of the FFTW interface is where the library is
> expected to stay.  I may try to add a "make install" if you need it.

Of course, I would appreciate greatly. But I do not wish to 
hassle you more than I have already. I am already idebted to 
you for writing the fftw package.

> Do not hesitate to report further difficulties with the installation
> -- I will try to write a quick INSTALL file from your remarks.

This is definitely worth the effort. But, it is such a pity 
that nothing can be done to rendere the installation 
painless until the new release is out.

Thank you very much.

Alex

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] Signal analysis

[Berke Durak]

On Mon, Jun 03, 2002 at 09:19:19PM +0200, Alessandro Baretta wrote:
> Hello.
> 
> I need some advice on a project I will have to start in a 
> few days. I need to write code to perform spectral analysis 
> of digital signals. 
[...]
> Are there any optimized FFT (and signal analysis) 
> libraries for Ocaml, or should I code the whole thing from 
> scratch?

Hello,

FFTW is really fast, but i've had trouble binding to it, so I ripped
the FFT code from Scheme's SLIB. The result is a small program that
displays in pseudo colour the spectrogram of a 16-bit raw audio
file, using Bigarray and Graphics :

	http://perso.wanadoo.fr/berke/caml/

BTW, couldn't someone hack FFTW to generate Caml code ? I mean, what's
the point of using C ?
--
Berke
-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] Signal analysis

[Markus Mottl]

On Mon, 03 Jun 2002, Berke Durak wrote:
> FFTW is really fast, but i've had trouble binding to it, so I ripped
> the FFT code from Scheme's SLIB. The result is a small program that
> displays in pseudo colour the spectrogram of a 16-bit raw audio
> file, using Bigarray and Graphics :
> 
> 	http://perso.wanadoo.fr/berke/caml/
> 
> BTW, couldn't someone hack FFTW to generate Caml code ? I mean, what's
> the point of using C ?

Interfacing to the FFTW-library is indeed awful, because of many options
that interact in fancy ways. Making a half-descent OCaml-library for
FFT-routines using the FFTW would probably take a couple of days full-time
(too long for me, I am more into symbolic stuff ;).

You can grab my unfinished attempt from here:

  http://www.oefai.at/~markus/tmp

Regards,
Markus Mottl

-- 
Markus Mottl                                             markus@oefai.at
Austrian Research Institute
for Artificial Intelligence                  http://www.oefai.at/~markus
-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners