论文标题
关于对抗性例子的非侵入性语音质量模型的鲁棒性
On the robustness of non-intrusive speech quality model by adversarial examples
论文作者
论文摘要
最近已经显示，基于深度学习的模型对语音质量预测有效，并且可以从各种角度胜过传统指标。尽管网络模型有可能成为复杂的人类听力感知的替代品，但它们可能在预测中包含不稳定性。这项工作表明，深层的语音质量预测因子可能容易受到对抗性扰动的影响，在这种情况下，与语音输入相比，通过不明显的扰动（$ -30 $ -30 $ dB）可以大大更改预测。除了揭示深层语音质量预测因子的脆弱性外，我们还进一步探索并确认了对抗性训练的生存能力，以增强模型的鲁棒性。
It has been shown recently that deep learning based models are effective on speech quality prediction and could outperform traditional metrics in various perspectives. Although network models have potential to be a surrogate for complex human hearing perception, they may contain instabilities in predictions. This work shows that deep speech quality predictors can be vulnerable to adversarial perturbations, where the prediction can be changed drastically by unnoticeable perturbations as small as $-30$ dB compared with speech inputs. In addition to exposing the vulnerability of deep speech quality predictors, we further explore and confirm the viability of adversarial training for strengthening robustness of models.
