Figure 4 - uploaded by Nils Peters
Content may be subject to copyright.
Contexts in source publication
Context 1
... we compare the PESQ computations as well as PESQ estimates by torchaudio-SQUIM in Figure 4. ...
Context 2
... systems, e.g., am-nsf-spk, showed an unexpected bimodal distribution that is not explained by the gender or the dataset, which needs further investigations. The SQUIM estimates for the reference signals, depicted by the right-most violin plot in the Figure 4, again show a bimodal behavior. ...
Context 3
... we compare the PESQ computations as well as PESQ estimates by torchaudio-SQUIM in Figure 4. ...
Context 4
... systems, e.g., am-nsf-spk, showed an unexpected bimodal distribution that is not explained by the gender or the dataset, which needs further investigations. The SQUIM estimates for the reference signals, depicted by the right-most violin plot in the Figure 4, again show a bimodal behavior. ...
Similar publications
Citations
... Considering prior research and VPC results [11], [12], it is likely that evaluation systems for voice anonymization algorithms also exhibit bias. Yet, unbiased evaluation is crucial to ensure that voice anonymization techniques perform well for any speaker subgroup. ...
... VPC evaluation framework contains only clean speech recordings and does not reflect real-world use cases. While there has been investigations of the reproduction capabilities, e.g., as in [2], this was also limited to clean acoustic conditions. As a result, in this work, we investigate the robustness of the VPC Baseline B1 under different noise conditions. ...
