[mary-users] EHMM Labelling segfaults

[Ingmar Steiner]

Dear Sriram,

On 30.03.2012 14:34, Sriram Shankar wrote:
> Hi,
>
> I am trying to run the voice import tool in openMARY and import an
> LPC-based diphone voice (cmu_us_kal_diphone) with a few phones of my
> own added.

The voice import tools in Mary 4 support building unit-selection and 
HMM-based voices, but AFAIA you cannot create diphone voices with them.

>
> When I try to run the EHMM labeller via the voice import tool, it
> crashes. I am attaching the crash report with this mail.
>
> My questions:
>
> 1. Any help in diagnosing the crash report?

It's the FeatureExtraction binary in EHMM that crashed. I believe you 
compiled it yourself from the sources in 
MARYBASE/lib/external/ehmm.tar.gz. You'll have to look there to debug 
what went wrong; it has nothing to do with Mary itself.

> 2. Since the LPC-based voice that I mentioned was running in Festival,
> do I need to go through all the tools in the voice import toolbox or
> is the voice ready for openMARY?
> If it is already ready, how do I start to use this voice?

That's an interesting question. Mary 4 uses the mbrola binaries for 
diphone synthesis. If you can get your Festival voice working with 
mbrola, it should be possible to load it in Mary, provided that the 
required files are in the correct format and found in the expected places.

Have a look at the package contents of a diphone voice installed with 
Mary, such as MARYBASE/download/mary-mbrola-us1-4.0.0.zip, specifically 
the voice config file and the files under MARYBASE/lib.

If you succeed in converting your diphone voice to the format required 
by Mary, please share your experience, since I'm sure it would be 
interesting for other users.

> 3. The crash report mentions that the core has been dumped, but I am
> not able to see where. Where would the core be dumped?

I'm afraid a discussion of C debugging is out of scope of this mailing 
list. Feel free to search the web.

Best wishes,

-Ingmar

P.S. It seems that emails from you are flagged with a very high (more 
than 80%) spam probability. Perhaps you should look into that.

>
> Regards,
> Sriram Shankar.
>
>
>
> _______________________________________________
> Mary-users mailing list
> Mary-users at dfki.de
> http://www.dfki.de/mailman/cgi-bin/listinfo/mary-users

-- 
Ingmar Steiner
Postdoctoral Researcher

LORIA Speech Group, Nancy, France
National Institute for Research in
Computer Science and Control (INRIA)