语音评判标准

Mel Cepstral Distortion ( M C D K \mathrm{MCD}_{K} MCDK)

M C D K = 1 T ∑ t = 0 T − 1 ∑ k = 1 K ( c t , k − c t , k ′ ) 2 \mathrm{MCD}_{K}=\frac{1}{T} \sum_{t=0}^{T-1} \sqrt{\sum_{k=1}^{K}\left(c_{t, k}-c_{t, k}^{\prime}\right)^{2}} MCDK=T1t=0T1k=1K(ct,kct,k)2

Gross Pitch Error (GPE)

G P E = ∑ t 1 [ ∣ p t − p t ′ ∣ > 0.2 p t ] 1 [ v t ] 1 [ v t ′ ] ∑ t 1 [ v t ] 1 [ v t ′ ] \mathrm{GPE}=\frac{\sum_{t} \mathbb{1}\left[\left|p_{t}-p_{t}^{\prime}\right|>0.2 p_{t}\right] \mathbb{1}\left[v_{t}\right] \mathbb{1}\left[v_{t}^{\prime}\right]}{\sum_{t} \mathbb{1}\left[v_{t}\right] \mathbb{1}\left[v_{t}^{\prime}\right]} GPE=t1[vt]1[vt]t1[ptpt>0.2pt]1[vt]1[vt]

Voicing Decision Error (VDE)

V D E = ∑ t = 0 T − 1 1 [ v t ≠ v t ′ ] T \mathrm{VDE}=\frac{\sum_{t=0}^{T-1} \mathbb{1}\left[v_{t} \neq v_{t}^{\prime}\right]}{T} VDE=Tt=0T11[vt=vt]

F0 Frame Error (FFE)

∑ t = 0 T − 1 1 [ ∣ p t − p t ′ ∣ > 0.2 p t ] 1 [ v t ] 1 [ v t ′ ] + 1 [ v t ≠ v t ′ ] T \frac{\sum_{t=0}^{T-1} \mathbb{1}\left[\left|p_{t}-p_{t}^{\prime}\right|>0.2 p_{t}\right] \mathbb{1}\left[v_{t}\right] \mathbb{1}\left[v_{t}^{\prime}\right]+\mathbb{1}\left[v_{t} \neq v_{t}^{\prime}\right]}{T} Tt=0T11[ptpt>0.2pt]1[vt]1[vt]+1[vt=vt]

posted @ 2020-10-16 13:37  赫凯  阅读(53)  评论(0)    收藏  举报