0
$\begingroup$

According to https://www.ntnu.edu/documents/1001201110/1266017954/DAFx-15_submission_43_v2.pdf, the audio processing toolbox Essentia includes the 100% of MPEG-7 low-level audio descriptors. However I have been trying to use it to extract them from a .wav file, and I can't find their names in Essentia's documentation.

My question is: Does Essentia actually include those descriptors out of the box (I mean, as a function, as it does with the MFCCs for instance), but maybe with different names than the originals? Or is it that it only includes "general tools" (as Spectrum extractors) to help with the actual extraction of those features?

In either case, could someone please explain here (or refer to a place where it is explained) how to get the seventeen MPEG-7 low-level descriptors USING ESSENTIA?

$\endgroup$
  • $\begingroup$ Indeed, take a look here. $\endgroup$ – jojek Mar 1 '18 at 12:59
  • $\begingroup$ I did, and they aren't mentioned anywhere in the page $\endgroup$ – Eddy Almuiña Mar 1 '18 at 22:18
  • $\begingroup$ Really? Which ones are missing? $\endgroup$ – jojek Mar 2 '18 at 9:02
  • $\begingroup$ audio waveform, audio spectrum envelope, audio spectrum centroid, audio spectrum spread, audio spectrum flatness... And those are just the ones I have had to implement by myself so far. $\endgroup$ – Eddy Almuiña Mar 2 '18 at 10:16
  • $\begingroup$ I know there are functions similar to those, but the thing is their results are not in the log-frequency bands specified in the MPEG-7 standard. So for that purpose they can't be directly employed. $\endgroup$ – Eddy Almuiña Mar 2 '18 at 10:18
0
$\begingroup$

Looking at the Essentia documentation, the MPEG-7 isn't listed directly as a function from where you can extract features.

The audio descriptors in the MPEG-7 standard are listed here. I think you will have to take use various functions to create your own feature vector. Hope this helps!

$\endgroup$
  • $\begingroup$ I think the same... The reason why I made this question is because according to the mentioned paper, Essentia does contain the 100% of MPEG-7 audio low-level descriptors, which to me seems to be inaccurate. $\endgroup$ – Eddy Almuiña Mar 2 '18 at 6:39
  • $\begingroup$ I'm not familiar with the paper but maybe it does but they don't have one function for all the descriptors. You may have to use multiple feature extraction functions. $\endgroup$ – Somesh Ganesh Mar 2 '18 at 14:41
  • $\begingroup$ Of course I didn't expect to get the 17 descriptors from a single function. My point is there are several descriptors which aren't included (IN ANY POSSIBLE WAY) into the Essentia library $\endgroup$ – Eddy Almuiña Mar 5 '18 at 5:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.