Medix-AI/coqui-tts
Medix-AI
/
coqui-tts
mirror of https://github.com/coqui-ai/TTS.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

refactor(audio.processor): remove duplicate linear_to_mel

This commit is contained in:
Enno Hermann 2023-11-14 10:41:01 +01:00
parent 4fd5c46937
commit fd9d6d4b0f
1 changed files with 5 additions and 16 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
TTS/utils/audio/processor.py
View File

@ -8,7 +8,7 @@ import scipy.signal
import soundfile as sf import soundfile as sf
from TTS.tts.utils.helpers import StandardScaler from TTS.tts.utils.helpers import StandardScaler
from TTS.utils.audio.numpy_transforms import amp_to_db, compute_f0, db_to_amp, stft, griffin_lim from TTS.utils.audio.numpy_transforms import amp_to_db, compute_f0, db_to_amp, griffin_lim, spec_to_mel, stft
# pylint: disable=too-many-public-methods # pylint: disable=too-many-public-methods
@ -410,17 +410,6 @@ class AudioProcessor(object):
return scipy.signal.lfilter([1], [1, -self.preemphasis], x) return scipy.signal.lfilter([1], [1, -self.preemphasis], x)
### SPECTROGRAMs ### ### SPECTROGRAMs ###
def _linear_to_mel(self, spectrogram: np.ndarray) -> np.ndarray:
"""Project a full scale spectrogram to a melspectrogram.
Args:
spectrogram (np.ndarray): Full scale spectrogram.
Returns:
np.ndarray: Melspectrogram
"""
return np.dot(self.mel_basis, spectrogram)
def _mel_to_linear(self, mel_spec: np.ndarray) -> np.ndarray: def _mel_to_linear(self, mel_spec: np.ndarray) -> np.ndarray:
"""Convert a melspectrogram to full scale spectrogram.""" """Convert a melspectrogram to full scale spectrogram."""
return np.maximum(1e-10, np.dot(self.inv_mel_basis, mel_spec)) return np.maximum(1e-10, np.dot(self.inv_mel_basis, mel_spec))
@ -460,10 +449,10 @@ class AudioProcessor(object):
win_length=self.win_length, win_length=self.win_length,
pad_mode=self.stft_pad_mode, pad_mode=self.stft_pad_mode,
) )
S = spec_to_mel(spec=np.abs(D), mel_basis=self.mel_basis)
if self.do_amp_to_db_mel: if self.do_amp_to_db_mel:
S = amp_to_db(x=self._linear_to_mel(np.abs(D)), gain=self.spec_gain, base=self.base) S = amp_to_db(x=S, gain=self.spec_gain, base=self.base)
else:
S = self._linear_to_mel(np.abs(D))
return self.normalize(S).astype(np.float32) return self.normalize(S).astype(np.float32)
def inv_spectrogram(self, spectrogram: np.ndarray) -> np.ndarray: def inv_spectrogram(self, spectrogram: np.ndarray) -> np.ndarray:
@ -493,7 +482,7 @@ class AudioProcessor(object):
""" """
S = self.denormalize(linear_spec) S = self.denormalize(linear_spec)
S = db_to_amp(x=S, gain=self.spec_gain, base=self.base) S = db_to_amp(x=S, gain=self.spec_gain, base=self.base)
S = self._linear_to_mel(np.abs(S)) S = spec_to_mel(spec=np.abs(S), mel_basis=self.mel_basis)
S = amp_to_db(x=S, gain=self.spec_gain, base=self.base) S = amp_to_db(x=S, gain=self.spec_gain, base=self.base)
mel = self.normalize(S) mel = self.normalize(S)
return mel return mel