Robustness simulated recording condition

Overall scores

	w2v2-L-cat	hubert-L-cat	wavlm-cat	data2vec-cat
Overall Score	33.3% passed tests (2 passed / 4 failed).	66.7% passed tests (4 passed / 2 failed).	33.3% passed tests (2 passed / 4 failed).	33.3% passed tests (2 passed / 4 failed).

Percentage Unchanged Predictions Simulated Position

Threshold: 0.8

Data	Percentage Unchanged Predictions Simulated Position
	w2v2-L-cat	hubert-L-cat	wavlm-cat	data2vec-cat
emovo-1.2.1-emotion.test	0.75	0.78	0.79	0.72
imda-nsc-read-speech-balanced-2.6.0-headset	0.81	0.90	0.87	0.79
timit-1.4.1-files	0.81	0.88	0.87	0.83
mean	0.79	0.85	0.84	0.78

Percentage Unchanged Predictions Simulated Room

Threshold: 0.8

Data	Percentage Unchanged Predictions Simulated Room
	w2v2-L-cat	hubert-L-cat	wavlm-cat	data2vec-cat
emovo-1.2.1-emotion.test	0.74	0.76	0.74	0.70
imda-nsc-read-speech-balanced-2.6.0-headset	0.76	0.88	0.79	0.74
timit-1.4.1-files	0.77	0.87	0.80	0.81
mean	0.76	0.84	0.78	0.75

Visualization Simulated Position

Confusion Matrix showing the shift from the predictions of audio with a baseline simulated position to the predictions of audio with a different simulated position.

w2v2-L-cat	hubert-L-cat	wavlm-cat	data2vec-cat