Representation Loss Minimization with Randomized Selection Strategy for Efficient Environmental Fake Audio Detection
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_ | 1866917784500305920 |
|---|---|
| author | Phukan, Orchid Chetia Girish Akhtar, Mohd Mujtaba Behera, Swarup Ranjan Choudhury, Nitin Buduru, Arun Balaji Sharma, Rajesh Prasanna, S. R Mahadeva |
| author_facet | Phukan, Orchid Chetia Girish Akhtar, Mohd Mujtaba Behera, Swarup Ranjan Choudhury, Nitin Buduru, Arun Balaji Sharma, Rajesh Prasanna, S. R Mahadeva |
| contents | The adaptation of foundation models has significantly advanced environmental audio deepfake detection (EADD), a rapidly growing area of research. These models are typically fine-tuned or utilized in their frozen states for downstream tasks. However, the dimensionality of their representations can substantially lead to a high parameter count of downstream models, leading to higher computational demands. So, a general way is to compress these representations by leveraging state-of-the-art (SOTA) unsupervised dimensionality reduction techniques (PCA, SVD, KPCA, GRP) for efficient EADD. However, with the application of such techniques, we observe a drop in performance. So in this paper, we show that representation vectors contain redundant information, and randomly selecting 40-50% of representation values and building downstream models on it preserves or sometimes even improves performance. We show that such random selection preserves more performance than the SOTA dimensionality reduction techniques while reducing model parameters and inference time by almost over half. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2409_15767 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Representation Loss Minimization with Randomized Selection Strategy for Efficient Environmental Fake Audio Detection Phukan, Orchid Chetia Girish Akhtar, Mohd Mujtaba Behera, Swarup Ranjan Choudhury, Nitin Buduru, Arun Balaji Sharma, Rajesh Prasanna, S. R Mahadeva Audio and Speech Processing Sound 68T45 I.2.7 The adaptation of foundation models has significantly advanced environmental audio deepfake detection (EADD), a rapidly growing area of research. These models are typically fine-tuned or utilized in their frozen states for downstream tasks. However, the dimensionality of their representations can substantially lead to a high parameter count of downstream models, leading to higher computational demands. So, a general way is to compress these representations by leveraging state-of-the-art (SOTA) unsupervised dimensionality reduction techniques (PCA, SVD, KPCA, GRP) for efficient EADD. However, with the application of such techniques, we observe a drop in performance. So in this paper, we show that representation vectors contain redundant information, and randomly selecting 40-50% of representation values and building downstream models on it preserves or sometimes even improves performance. We show that such random selection preserves more performance than the SOTA dimensionality reduction techniques while reducing model parameters and inference time by almost over half. |
| title | Representation Loss Minimization with Randomized Selection Strategy for Efficient Environmental Fake Audio Detection |
| topic | Audio and Speech Processing Sound 68T45 I.2.7 |
| url | https://arxiv.org/abs/2409.15767 |