Nonlinear Regularization Decoding Method for Speech Recognition
Sensors · 2024
DOI: 10.3390/s24123846 ↗Kurz gefasst
Fachartikel in der Zeitschrift Sensors, 2024. Verfasst von Jiang Zhang und 3 weiteren Autor:innen. Beteiligte Einrichtung: College of Computer Science and Technology, Xinjiang University, Urumqi 830017, China. Hauptthemen laut MEDLINE: Speech Recognition Software, Algorithms. Die Arbeit wurde bisher 8-mal zitiert (Crossref) und nennt 44 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: the Tianshan Excellence Program Project of Xinjiang Uygur Autonomous Region, the Central Government Guides Local Science and Technology Development Fund Projects, the Graduate Research Innovation Project of Xinjiang Uygur Autonomous Region.
Abstract (Originalsprache)
Existing end-to-end speech recognition methods typically employ hybrid decoders based on CTC and Transformer. However, the issue of error accumulation in these hybrid decoders hinders further improvements in accuracy. Additionally, most existing models are built upon Transformer architecture, which tends to be complex and unfriendly to small datasets. Hence, we propose a Nonlinear Regularization Decoding Method for Speech Recognition. Firstly, we introduce the nonlinear Transformer decoder, breaking away from traditional left-to-right or right-to-left decoding orders and enabling associations between any characters, mitigating the limitations of Transformer architectures on small datasets. Secondly, we propose a novel regularization attention module to optimize the attention score matrix, reducing the impact of early errors on later outputs. Finally, we introduce the tiny model to address the challenge of overly large model parameters. The experimental results indicate that our model demonstrates good performance. Compared to the baseline, our model achieves recognition improvements of 0.12%, 0.54%, 0.51%, and 1.2% on the Aishell1, Primewords, Free ST Chinese Corpus, and Common Voice 16.1 datasets of Uyghur, respectively.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- College of Computer Science and Technology, Xinjiang University, Urumqi 830017, China.
ORCID-Profile: Liejun Wang, Yinfeng Yu
Förderung
- the Tianshan Excellence Program Project of Xinjiang Uygur Autonomous Region
- the Central Government Guides Local Science and Technology Development Fund Projects
- the Graduate Research Innovation Project of Xinjiang Uygur Autonomous Region
- Tianshan Excellence Program Project of Xinjiang Uygur Autonomous Region
- Central Government Guides Local Science and Technology Development Fund Projects
- Graduate Research Innovation Project of Xinjiang Uygur Autonomous Region
Zitieren
APA
Zhang, J., Wang, L., Yu, Y., & Xu, M. (2024). Nonlinear Regularization Decoding Method for Speech Recognition. Sensors. https://doi.org/10.3390/s24123846
Vancouver
Zhang J, Wang L, Yu Y, Xu M. Nonlinear Regularization Decoding Method for Speech Recognition. Sensors. 2024. doi:10.3390/s24123846
BibTeX
@article{Zhang2024,
author = {Zhang, Jiang and Wang, Liejun and Yu, Yinfeng and Xu, Miaomiao},
title = {Nonlinear Regularization Decoding Method for Speech Recognition},
journal = {Sensors},
year = {2024},
doi = {10.3390/s24123846}
}
EUVIMED