Publication Details

Auxiliary Loss Function for Target Speech Extraction and Recognition with Weak Supervision Based on Speaker Characteristics

ŽMOLÍKOVÁ, K.; DELCROIX, M.; RAJ, D.; WATANABE, S.; ČERNOCKÝ, J. Auxiliary Loss Function for Target Speech Extraction and Recognition with Weak Supervision Based on Speaker Characteristics. In Proceedings of 2021 Interspeech. Proceedings of Interspeech. Brno: International Speech Communication Association, 2021. p. 1464-1468. ISSN: 1990-9772.
Czech title
Pomocná ztrátová funkce pro extrakci a rozpoznávání řeči cílového mluvčího se slabou supervizí založenou na charakteristice mluvčího
Type
conference paper
Language
English
Authors
Žmolíková Kateřina, Ing., Ph.D. (FIT)
Delcroix Marc
RAJ, D.
Watanabe Shinji
Černocký Jan, prof. Dr. Ing. (DCGM)
URL
Keywords

Target speech extraction, SpeakerBeam, Weakly supervised loss, Long recordings

Abstract

Automatic speech recognition systems deteriorate in presence of overlapped
speech. A popular approach to alleviate this is target speech extraction. The
extraction system is usually trained with a loss function measuring the
discrepancy between the estimated and the reference target speech. This often
leads to distortions to the target signal which is detrimental to the recognition
accuracy. Additionally, it is necessary to have the strong supervision provided
by parallel data consisting of speech mixtures and single-speaker signals. We
propose an auxiliary loss function for retraining the target speech extraction.
It is composed of two parts: first, a speaker identity loss, forcing the
estimated speech to have correct speaker characteristics, and second, a mixture
consistency loss, making the extracted sources sum back to the original mixture.
The only supervision required for the proposed loss is speaker characteristics
obtained from several segments spoken by the target speaker. Such weak
supervision makes the loss suitable for adapting the system directly on real
recordings. We show that the proposed loss yields signals more suitable for
speech recognition and further, we can gain additional improvements by adaptation
to target data. Overall, we can reduce the word error rate on LibriCSS dataset
from 27.4% to 24.0%.

Published
2021
Pages
1464–1468
Journal
Proceedings of Interspeech, vol. 2021, no. 8, ISSN 1990-9772
Proceedings
Proceedings of 2021 Interspeech
Conference
Interspeech Conference, Brno, CZ
Publisher
International Speech Communication Association
Place
Brno
DOI
UT WoS
000841879501116
EID Scopus
BibTeX
@inproceedings{BUT175837,
  author="ŽMOLÍKOVÁ, K. and DELCROIX, M. and RAJ, D. and WATANABE, S. and ČERNOCKÝ, J.",
  title="Auxiliary Loss Function for Target Speech Extraction and Recognition with Weak Supervision Based on Speaker Characteristics",
  booktitle="Proceedings of 2021 Interspeech",
  year="2021",
  journal="Proceedings of Interspeech",
  volume="2021",
  number="8",
  pages="1464--1468",
  publisher="International Speech Communication Association",
  address="Brno",
  doi="10.21437/Interspeech.2021-986",
  issn="1990-9772",
  url="https://www.isca-speech.org/archive/interspeech_2021/zmolikova21_interspeech.html"
}
Files
Back to top