Emotion Recognition from Speech Using wav2vec 2.0 Embeddings

Leonardo Pepino, Pablo Riera, Luciana Ferrer · 2021

Read Emotion Recognition from Speech Using wav2vec 2.0 Embeddings by Leonardo Pepino, Pablo Riera, Luciana Ferrer (2021) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/openalex-w3144073776.txt.

Description

Leonardo Pepino, Pablo Riera, and Luciana Ferrer presented this paper at Interspeech in 2021. It tackles emotion recognition from speech, teaching computers to hear anger, joy, or sadness in a voice. Labeled emotion recordings are scarce, which starves the deep learning models that work best.

Their fix is transfer learning. A large speech model, wav2vec 2.0, first learns the sound of speech from unlabeled audio. Its internal features then feed a small network trained for emotion, blending several layers with learned weights. On two standard test sets the method beat earlier published results. The study shows borrowed knowledge can stand in for missing data in speech technology.

Catalog record

CREATORLeonardo Pepino, Pablo Riera, Luciana Ferrer
YEAR2021
COLLECTIONPAPERS
FORMATTXT
FILE SIZE32 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEOpenAlex / Interspeech (USB)

Source: https://doi.org/10.21437/interspeech.2021-703

Browse the full library · About the Tower of Babel