EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

A CTC-Based Speech Recognition Network Fusing Local Convolution and Global Attention

Huijuan Hu, Chenyang Tang, Ping Tan, He Xu iD

Sensors · 16.03.2026 · 26 (6) · 1865

DOI: 10.3390/s26061865 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 16.03.2026. Verfasst von Huijuan Hu und 3 weiteren Autor:innen. Beteiligte Einrichtung: School of Computer Science, Nanjing University of Posts and Telecommunications, Nanjing 210023, China. Die Arbeit nennt 36 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Postgraduate Research and Practice Innovation Program of Jiangsu Province.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Integrating wav2vec 2.0 with Connectionist Temporal Classification (CTC) for automatic speech recognition (ASR) often involves a trade-off between capturing global semantic consistency and maintaining local feature discriminability. This study proposes DBA-wav2vec 2.0, an architecture designed to manage these modeling requirements by decoupling temporal modeling into parallel local and global streams at the encoder–decoder interface. Depthwise separable convolutions are utilized to capture local acoustic structures, while a self-attention path is retained for long-range dependencies. A task-aware gating mechanism is introduced to integrate these heterogeneous features. By adjusting fusion weights based on acoustic input characteristics, the gate facilitates the refinement of posterior probability distributions, leading to more distinct alignment points. Experimental results on AISHELL-1 and ST-CMDS datasets show relative Character Error Rate (CER) reductions of 6.4% and 7.4%, respectively, compared to a baseline wav2vec 2.0 model. Further evaluations under varying speaking rates demonstrate a 15.3% relative improvement in fast-speech scenarios, suggesting that structural adaptation at the decoding interface can enhance the robustness of CTC-based systems against temporal variations.

Schlagwörter

Stichwörter der Autor:innen

CTC alignmentautomatic speech recognitiondual-branch architectureinformation fusiontask-aware gatingwav2vec 2.0

Autor:innen und Institutionen

  • School of Computer Science, Nanjing University of Posts and Telecommunications, Nanjing 210023, China.
  • School of Business, Tongda College of Nanjing University of Posts and Telecommunications, Yangzhou 225127, China.

ORCID-Profile: He Xu

Förderung

  • Postgraduate Research and Practice Innovation Program of Jiangsu Province

Zitieren

APA

Hu, H., Tang, C., Tan, P., & Xu, H. (2026). A CTC-Based Speech Recognition Network Fusing Local Convolution and Global Attention. Sensors, 26(6), 1865. https://doi.org/10.3390/s26061865

Vancouver

Hu H, Tang C, Tan P, Xu H. A CTC-Based Speech Recognition Network Fusing Local Convolution and Global Attention. Sensors. 2026;26(6):1865. doi:10.3390/s26061865

BibTeX
@article{Hu2026,
  author = {Hu, Huijuan and Tang, Chenyang and Tan, Ping and Xu, He},
  title = {A CTC-Based Speech Recognition Network Fusing Local Convolution and Global Attention},
  journal = {Sensors},
  year = {2026},
  volume = {26},
  number = {6},
  pages = {1865},
  doi = {10.3390/s26061865}
}
Kostenlos starten →