MPSA-Conformer-CTC/Attention: A High-Accuracy, Low-Complexity End-to-End Approach for Tibetan Speech Recognition
Sensors · 2024
DOI: 10.3390/s24216824 ↗Kurz gefasst
Fachartikel in der Zeitschrift Sensors, 2024. Verfasst von Changlin Wu und 3 weiteren Autor:innen. Beteiligte Einrichtung: School of Mechanical and Electrical Engineering, Huainan Normal University, Huainan 232001, China. Hauptthemen laut MEDLINE: Algorithms, Speech Recognition Software. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 44 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: the Training Programme Foundation for the Talents by Universities of Anhui Province, the State Key Laboratory Open Fund of Mining Response and Disaster Prevention , Control in Deep Coal Mines, the Fundamental Research Funds for the Central Universities.
Abstract (Originalsprache)
This study addresses the challenges of low accuracy and high computational demands in Tibetan speech recognition by investigating the application of end-to-end networks. We propose a decoding strategy that integrates Connectionist Temporal Classification (CTC) and Attention mechanisms, capitalizing on the benefits of automatic alignment and attention weight extraction. The Conformer architecture is utilized as the encoder, leading to the development of the Conformer-CTC/Attention model. This model first extracts global features from the speech signal using the Conformer, followed by joint decoding of these features through CTC and Attention mechanisms. To mitigate convergence issues during training, particularly with longer input feature sequences, we introduce a Probabilistic Sparse Attention mechanism within the joint CTC/Attention framework. Additionally, we implement a maximum entropy optimization algorithm for CTC, effectively addressing challenges such as increased path counts, spike distributions, and local optima during training. We designate the proposed method as the MaxEnt-Optimized Probabilistic Sparse Attention Conformer-CTC/Attention Model (MPSA-Conformer-CTC/Attention). Experimental results indicate that our improved model achieves a word error rate reduction of 10.68% and 9.57% on self-constructed and open-source Tibetan datasets, respectively, compared to the baseline model. Furthermore, the enhanced model not only reduces memory consumption and training time but also improves generalization capability and accuracy.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- School of Mechanical and Electrical Engineering, Huainan Normal University, Huainan 232001, China.
- Human-Computer Collaborative Robot Joint Laboratory of Anhui Province, Hefei 230002, China.
Förderung
- the Training Programme Foundation for the Talents by Universities of Anhui Province
- the State Key Laboratory Open Fund of Mining Response and Disaster Prevention , Control in Deep Coal Mines
- the Fundamental Research Funds for the Central Universities
- State Key Laboratory Open Fund of Mining Response and Disaster Prevention, Control in Deep Coal Mines
- Training Programme Foundation for the Talents by Universities of Anhui Province
- Fundamental Research Funds for the Central Universities
Zitieren
APA
Wu, C., Sun, H., Huang, K., & Wu, L. (2024). MPSA-Conformer-CTC/Attention: A High-Accuracy, Low-Complexity End-to-End Approach for Tibetan Speech Recognition. Sensors. https://doi.org/10.3390/s24216824
Vancouver
Wu C, Sun H, Huang K, Wu L. MPSA-Conformer-CTC/Attention: A High-Accuracy, Low-Complexity End-to-End Approach for Tibetan Speech Recognition. Sensors. 2024. doi:10.3390/s24216824
BibTeX
@article{Wu2024,
author = {Wu, Changlin and Sun, Huihui and Huang, Kaifeng and Wu, Long},
title = {MPSA-Conformer-CTC/Attention: A High-Accuracy, Low-Complexity End-to-End Approach for Tibetan Speech Recognition},
journal = {Sensors},
year = {2024},
doi = {10.3390/s24216824}
}
EUVIMED