CODEWITHSUNDEEP
audiofeature-extraction
user3140427
user3140427

Reputation: 19

Extracting audio features from real-time input, using OpenSMILE

I can extract features from audio files stored in system, using OpenSMILE. But I need to extract features in real-time, using the same OpenSMILE. Is it possible, without saving the recorded audio file?

Upvotes: 1

Views: 4348

Answers (3)

Florian
Florian

Reputation: 126

You might also need a cTurnDetector component (see the SIG'MM tutorial referenced at http://opensmile.audeering.com/) and a VAD to enable segmentation of your live input for some features (functionals).

Also check out: How to create custom config files in OpenSMILE to see how you can customize configuration files and set the buffer sizes correctly (in live mode it is important to use ring buffers to avoid growing memory consumption).

Upvotes: 1

tugce.kolcu
tugce.kolcu

Reputation: 55

It works properly if you build the openSMILE with portAudio :)

According to manual:

... A live feature extraction example is provided, which captures live audio and extracts prosodic features (pitch and loudness contours) from the input. The features are saved to a CSV le. To use this con guration, type:

SMILExtract -􀀀C config/liveProsodyAcf.conf

Upvotes: 0

AShelly
AShelly

Reputation: 35540

It seems it may support it. From the manual

To check if your SMILExtract binary supports live audio recording and playback, type: SMILExtract -H cPortaudio

and

The components cPortaudioSource and cPortaudioSink can be used as replacements for cWaveSource and cWaveSink. They produce/expect data in the same format as the wave components.

Hooray for internet searching...

Upvotes: 2

Related Questions