How to use custom feature for speaker recongition

[Quazi Marufur Rahman]

Hi,

Is there any way to use custom features, bypassing the default extraction and preprocessing steps of bob.bio.spear? Actually I have extracted some features using customized algorithm and want to use those for speaker recognition. What I am looking for, is a place in source code where I can plug my algorithm to provide necessary features for a particular wav file.

Thanks 

[Tiago Freitas Pereira]

Hi Quazi,

If your feature are in the HDF5 format (or in any format that bob.io.base.load accepts), you can by pass it just setting the flags.

   --extracted-directory <YOUR feature directory>
   --skip-extraction --skip-preprocessing

Cheers

Tiago

--
-- You received this message because you are subscribed to the Google Groups bob-devel group. To post to this group, send email to bob-...@googlegroups.com. To unsubscribe from this group, send email to bob-devel+...@googlegroups.com. For more options, visit this group at https://groups.google.com/d/forum/bob-devel or directly the project website at http://idiap.github.com/bob/
---
You received this message because you are subscribed to the Google Groups "bob-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bob-devel+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Tiago

[Manuel Günther]

If your files are not in the desired format (or you don't have them in files yet at all), you might want to write your own feature extractor. Please read http://pythonhosted.org/bob.bio.base/index.html (particularly the section about the Implementation Details) for more details. Let us know if you have any difficulties.

Am Donnerstag, 10. März 2016 10:08:11 UTC-7 schrieb Tiago Freitas Pereira:

Hi Quazi,

If your feature are in the HDF5 format (or in any format that bob.io.base.load accepts), you can by pass it just setting the flags.

   --extracted-directory <YOUR feature directory>
   --skip-extraction --skip-preprocessing

Cheers

Tiago

On Thu, Mar 10, 2016 at 11:59 AM, Quazi Marufur Rahman wrote:

Hi,

Is there any way to use custom features, bypassing the default extraction and preprocessing steps of bob.bio.spear? Actually I have extracted some features using customized algorithm and want to use those for speaker recognition. What I am looking for, is a place in source code where I can plug my algorithm to provide necessary features for a particular wav file.

Thanks 

--

Tiago

[Quazi Marufur Rahman]

Thanks for your suggestions. 

I have directly edited extractor.py to use the customized feature.

--

-- You received this message because you are subscribed to the Google Groups bob-devel group. To post to this group, send email to bob-...@googlegroups.com. To unsubscribe from this group, send email to bob-devel+...@googlegroups.com. For more options, visit this group at https://groups.google.com/d/forum/bob-devel or directly the project website at http://idiap.github.com/bob/
---

You received this message because you are subscribed to a topic in the Google Groups "bob-devel" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/bob-devel/aF07RAL7dzI/unsubscribe.
To unsubscribe from this group and all its topics, send an email to bob-devel+...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--

Maruf