opensmile 1.0.0

Feature sets

Currently, three standard sets are supported. ComParE 2016 is the largest with more than 6k features. The smaller sets GeMAPS and eGeMAPS come in two variants v01a and v01b. We suggest to use the newer version unless backward compatibility with the original papers is desired.

Each feature set can be extracted on three levels:

• Low-level descriptors (LDD)

• LLDs with deltas

• Functionals

The following table lists the number of features for each set and level.

Code example

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')

License

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 info@audeering.com (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

Citing

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.