Skip to main content
. Author manuscript; available in PMC: 2022 Nov 4.
Published in final edited form as: Interspeech. 2022 Sep;2022:3338–3342. doi: 10.21437/interspeech.2022-10798

Table 5:

J-ratios and F1-Avg scores for three input features with and without adversarial learning on CONVERGE data. ‘+Adv’ denotes adversarial training

Input Features J-Ratio F1-Avg
Mel-spec 42.66 0.879
Mel-spec+Adv 40.61 0.890
Raw-audio 302.3 0.829
Raw-audio+Adv 74.64 0.857
XLSR-Mandarin 75.54 0.912
XLSR-Mandarin+Adv 71.62 0.915