BanglaRobustNet: A Hybrid Denoising-Attention Architecture for Robust Bangla Speech Recognition
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866911396959092736 |
|---|---|
| author | Ridoy, Md Sazzadul Islam Zidney, Mubaswira Ibnat Akter, Sumi Rahman, Md. Aminur |
| author_facet | Ridoy, Md Sazzadul Islam Zidney, Mubaswira Ibnat Akter, Sumi Rahman, Md. Aminur |
| contents | Bangla, one of the most widely spoken languages, remains underrepresented in state-of-the-art automatic speech recognition (ASR) research, particularly under noisy and speaker-diverse conditions. This paper presents BanglaRobustNet, a hybrid denoising-attention framework built on Wav2Vec-BERT, designed to address these challenges. The architecture integrates a diffusion-based denoising module to suppress environmental noise while preserving Bangla-specific phonetic cues, and a contextual cross-attention module that conditions recognition on speaker embeddings for robustness across gender, age, and dialects. Trained end-to-end with a composite objective combining CTC loss, phonetic consistency, and speaker alignment, BanglaRobustNet achieves substantial reductions in word error rate (WER) and character error rate (CER) compared to Wav2Vec-BERT and Whisper baselines. Evaluations on Mozilla Common Voice Bangla and augmented noisy speech confirm the effectiveness of our approach, establishing BanglaRobustNet as a robust ASR system tailored to low-resource, noise-prone linguistic settings. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_17679 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | BanglaRobustNet: A Hybrid Denoising-Attention Architecture for Robust Bangla Speech Recognition Ridoy, Md Sazzadul Islam Zidney, Mubaswira Ibnat Akter, Sumi Rahman, Md. Aminur Sound Computation and Language Computer Vision and Pattern Recognition Machine Learning Audio and Speech Processing Bangla, one of the most widely spoken languages, remains underrepresented in state-of-the-art automatic speech recognition (ASR) research, particularly under noisy and speaker-diverse conditions. This paper presents BanglaRobustNet, a hybrid denoising-attention framework built on Wav2Vec-BERT, designed to address these challenges. The architecture integrates a diffusion-based denoising module to suppress environmental noise while preserving Bangla-specific phonetic cues, and a contextual cross-attention module that conditions recognition on speaker embeddings for robustness across gender, age, and dialects. Trained end-to-end with a composite objective combining CTC loss, phonetic consistency, and speaker alignment, BanglaRobustNet achieves substantial reductions in word error rate (WER) and character error rate (CER) compared to Wav2Vec-BERT and Whisper baselines. Evaluations on Mozilla Common Voice Bangla and augmented noisy speech confirm the effectiveness of our approach, establishing BanglaRobustNet as a robust ASR system tailored to low-resource, noise-prone linguistic settings. |
| title | BanglaRobustNet: A Hybrid Denoising-Attention Architecture for Robust Bangla Speech Recognition |
| topic | Sound Computation and Language Computer Vision and Pattern Recognition Machine Learning Audio and Speech Processing |
| url | https://arxiv.org/abs/2601.17679 |