search

audeering / opensmile-python

Python package for openSMILE
https://audeering.github.io/opensmile-python/
Other
240 stars 32 forks source link

How to convert LLDs into Functionals？ #81

Open Shen-JK opened 1 year ago

Shen-JK commented 1 year ago

Hi，

I don't have the original wav but only get LLDs. Is there a function that allows me to convert LLDs directly to Functionals?

Thanks!

ckarjadi commented 10 months ago

Hi Shen-JK,

Unfortunately, it seems to be tricky.

I asked info@audeering.com this question and their response was:

Hi Cody,

theoretically it is possible to read CSVs with LLDs and compute functionals. However, we have never implemented a config file for this case, as the metadata (framesizes, etc.) gets lost in the CSV and it was just the most convenient to run on wav again.

If you want to explore this option, you could start by creating a custom config file with a csv reader, and the functionals components from the respective configuration file. The LLDs are stored in quite a few different levels though, and if you read a CSV, they would be in one level only. So you would either need to split the features after reading the CSV based on their names, or you save each LLD level in a separate CSV. I would not be able to tell you, if in the end it would work, as there might be metadata like framesizes of sampling rate of the audio file missing, which some functionals might need to compute proper values.

Best,

Florian