An Exploration of Length Generalization in Transformer-Based Speech Enhancement
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866913393500225536 |
|---|---|
| author | Zhang, Qiquan Zhu, Hongxu Qian, Xinyuan Ambikairajah, Eliathamby Li, Haizhou |
| author_facet | Zhang, Qiquan Zhu, Hongxu Qian, Xinyuan Ambikairajah, Eliathamby Li, Haizhou |
| contents | The use of Transformer architectures has facilitated remarkable progress in speech enhancement. Training Transformers using substantially long speech utterances is often infeasible as self-attention suffers from quadratic complexity. It is a critical and unexplored challenge for a Transformer-based speech enhancement model to learn from short speech utterances and generalize to longer ones. In this paper, we conduct comprehensive experiments to explore the length generalization problem in speech enhancement with Transformer. Our findings first establish that position embedding provides an effective instrument to alleviate the impact of utterance length on Transformer-based speech enhancement. Specifically, we explore four different position embedding schemes to enable length generalization. The results confirm the superiority of relative position embeddings (RPEs) over absolute PE (APEs) in length generalization. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_11401 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | An Exploration of Length Generalization in Transformer-Based Speech Enhancement Zhang, Qiquan Zhu, Hongxu Qian, Xinyuan Ambikairajah, Eliathamby Li, Haizhou Audio and Speech Processing The use of Transformer architectures has facilitated remarkable progress in speech enhancement. Training Transformers using substantially long speech utterances is often infeasible as self-attention suffers from quadratic complexity. It is a critical and unexplored challenge for a Transformer-based speech enhancement model to learn from short speech utterances and generalize to longer ones. In this paper, we conduct comprehensive experiments to explore the length generalization problem in speech enhancement with Transformer. Our findings first establish that position embedding provides an effective instrument to alleviate the impact of utterance length on Transformer-based speech enhancement. Specifically, we explore four different position embedding schemes to enable length generalization. The results confirm the superiority of relative position embeddings (RPEs) over absolute PE (APEs) in length generalization. |
| title | An Exploration of Length Generalization in Transformer-Based Speech Enhancement |
| topic | Audio and Speech Processing |
| url | https://arxiv.org/abs/2406.11401 |