Table 5. Naturalness MOS and quality MOS... | Emerald Publishing

Table 5.

Naturalness MOS and quality MOS for each speaker. The ± indicates the 95% confidence intervals for the scores

	Naturalness MOS $↑$			Quality MOS $↑$
Ground truth	$3.70 \pm 0.11$			$3.70 \pm 0.10$
Speaker ID	TF (Chung et al., 2024)	Ours	SA2 (Evans et al., 2024)	TF (Chung et al., 2024)	Ours	SA2 (Evans et al., 2024)
m-01	$2.86 \pm 0.12$	$3.50 \pm 0.11$	$3.29 \pm 0.11$	$2.73 \pm 0.12$	$3.36 \pm 0.11$	$3.36 \pm 0.10$
m-02	$2.56 \pm 0.11$	$3.09 \pm 0.12$	$3.42 \pm 0.10$	$2.47 \pm 0.11$	$3.16 \pm 0.11$	$3.45 \pm 0.11$
m-03	$2.82 \pm 0.12$	$3.46 \pm 0.11$	$3.48 \pm 0.10$	$2.55 \pm 0.11$	$3.32 \pm 0.10$	$3.52 \pm 0.10$
m-04	$2.95 \pm 0.12$	$3.55 \pm 0.11$	$3.57 \pm 0.10$	$2.83 \pm 0.12$	$3.39 \pm 0.10$	$3.48 \pm 0.10$
m-05	$2.69 \pm 0.11$	$2.93 \pm 0.12$	$3.72 \pm 0.10$	$2.49 \pm 0.11$	$2.95 \pm 0.11$	$3.57 \pm 0.10$
f-01	$2.76 \pm 0.12$	$3.14 \pm 0.12$	$3.53 \pm 0.11$	$2.69 \pm 0.11$	$2.99 \pm 0.11$	$3.45 \pm 0.10$
Whole	$2.77 \pm 0.05$	$3.28 \pm 0.05$	$3.50 \pm 0.04$	$2.62 \pm 0.05$	$3.19 \pm 0.04$	$3.47 \pm 0.04$

	Naturalness MOS $↑$			Quality MOS $↑$
Ground truth	$3.70 \pm 0.11$			$3.70 \pm 0.10$
Speaker ID	TF (Chung et al., 2024)	Ours	SA2 (Evans et al., 2024)	TF (Chung et al., 2024)	Ours	SA2 (Evans et al., 2024)
m-01	$2.86 \pm 0.12$	$3.50 \pm 0.11$	$3.29 \pm 0.11$	$2.73 \pm 0.12$	$3.36 \pm 0.11$	$3.36 \pm 0.10$
m-02	$2.56 \pm 0.11$	$3.09 \pm 0.12$	$3.42 \pm 0.10$	$2.47 \pm 0.11$	$3.16 \pm 0.11$	$3.45 \pm 0.11$
m-03	$2.82 \pm 0.12$	$3.46 \pm 0.11$	$3.48 \pm 0.10$	$2.55 \pm 0.11$	$3.32 \pm 0.10$	$3.52 \pm 0.10$
m-04	$2.95 \pm 0.12$	$3.55 \pm 0.11$	$3.57 \pm 0.10$	$2.83 \pm 0.12$	$3.39 \pm 0.10$	$3.48 \pm 0.10$
m-05	$2.69 \pm 0.11$	$2.93 \pm 0.12$	$3.72 \pm 0.10$	$2.49 \pm 0.11$	$2.95 \pm 0.11$	$3.57 \pm 0.10$
f-01	$2.76 \pm 0.12$	$3.14 \pm 0.12$	$3.53 \pm 0.11$	$2.69 \pm 0.11$	$2.99 \pm 0.11$	$3.45 \pm 0.10$
Whole	$2.77 \pm 0.05$	$3.28 \pm 0.05$	$3.50 \pm 0.04$	$2.62 \pm 0.05$	$3.19 \pm 0.04$	$3.47 \pm 0.04$