Re: Q: speech toolbox?

[Pascal Fleury]

Jan Buckow wrote:

> Hi,
>
> I recently dicovered octave and would like to use it for short
> experiments. As I am working on speech research, I was wondering
> if someone has already developed a toolbox to deal with speech
> signals, e.g.
>
> - read/write audio

>

I have written some glue code to link libsndfile into octave. The advantage of
this (as to the supported formats of Octave) is that:
- libsndfile handles more file formats (on many platforms)
- my glue code enables you to read/write only a block of the input/output
file, so for sliding window analysis of very large signals, you do not need an
prohibitively high amount of RAM (you can also read the whole file if you want
to. It is a 5 lines Octave function)

Disadvantage is that it is fairly recent code I wrote, and it has no
tar/make/make install installation procedure, and it has a number of known
inefficiencies and a memory leak with no simple work-around (at least to me),
and it depends on an external library (I don't know if that is an issue for
Octave maintainers). I am happy to contribute it however.

For the library itself, please see http://www.zip.com.au/Šþerikd/libsndfile/

In the more or less long run, I could contribute this to octave, as an
optional package to be installed of the libsndfile is present. For the moment,
just ask me, I'll set you the code. I have no time to do extensive support
though, so you will be left on your own. (He, it's almost Christmas, and I
have a family :-)

And, yes, thank you John for the nice job you did ! What about the future of
Octave ? I will use it ! (no other comment)

Pascal

fleury.vcf
Description: Card for Pascal Fleury