Python interface for extracting openSMILE features.
$ pip install opensmile
Note
Only 64-bit Python is supported.
Currently, three standard sets are supported.
ComParE 2016 is the largest with more than 6k features.
The smaller sets GeMAPS and eGeMAPS
come in variants v01a, v01b and v02 (only eGeMAPS).
We suggest to use the latest version
unless backward compatibility with
the original papers is desired.
Each feature set can be extracted on two levels:
- Low-level descriptors (LDD)
- Functionals
For ComParE 2016 a third level is available:
- LLD deltas
Note
Pre v2.0.0 some LLDs of the GeMAPS family were incorrectly output as deltas. This was corrected with v2.0.0 and these features are now correctly returned as LLDs. Note that with v2.0.0 deltas are no longer available for the GeMAPS family.
The following table lists the number of features for each set and level.
| Name | #features |
|---|---|
| ComParE_2016 | 65 / 65 / 6373 |
| GeMAPSv01a | 18 / - / 62 |
| GeMAPSv01b | 18 / - / 62 |
| eGeMAPSv01a | 23 / - / 88 |
| eGeMAPSv01b | 23 / - / 88 |
| eGeMAPSv02 | 25 / - / 88 |
Note
Additional feature sets have been added by the community.
For a full list please see the documentation of opensmile.FeatureSet.
| Name | #features |
|---|---|
| ComParE_2016 | 65 / 65 / 6373 |
| GeMAPSv01a | 5 / 13 / 62 |
| GeMAPSv01b | 5 / 13 / 62 |
| eGeMAPSv01a | 10 / 13 / 88 |
| eGeMAPSv01b | 10 / 13 / 88 |
Code example, that extracts ComParE 2016 functionals from an audio file:
import opensmile
smile = opensmile.Smile(
feature_set=opensmile.FeatureSet.ComParE_2016,
feature_level=opensmile.FeatureLevel.Functionals,
)
y = smile.process_file('audio.wav')openSMILE follows a dual-licensing model. Since the main goal of the project is a widespread use of the software to facilitate research in the field of machine learning from audio-visual signals, the source code and binaries are freely available for private, research, and educational use under an open-source license (see LICENSE). It is not allowed to use the open-source version of openSMILE for any sort of commercial product. Fundamental research in companies, for example, is permitted, but if a product is the result of the research, we require you to buy a commercial development license. Contact us at [email protected] (or visit us at https://www.audeering.com) for more information.
Original authors: Florian Eyben, Felix Weninger, Martin Wöllmer, Björn Schuller
Copyright © 2008-2013, Institute for Human-Machine Communication, Technische Universität München, Germany
Copyright © 2013-2015, audEERING UG (haftungsbeschränkt)
Copyright © 2016-2020, audEERING GmbH
Please cite openSMILE in your publications by citing the following paper:
Florian Eyben, Martin Wöllmer, Björn Schuller: "openSMILE - The Munich Versatile and Fast Open-Source Audio Feature Extractor", Proc. ACM Multimedia (MM), ACM, Florence, Italy, ISBN 978-1-60558-933-6, pp. 1459-1462, 25.-29.10.2010.
