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

Remove min max mel freq

pull/10/head
Eren 2018-09-06 15:25:47 +02:00
parent 070156b631
commit 56c6d0cac8
2 changed files with 1 additions and 21 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
train.py
View File

@ -347,9 +347,7 @@ def main(args):
ref_level_db=c.ref_level_db, ref_level_db=c.ref_level_db,
num_freq=c.num_freq, num_freq=c.num_freq,
power=c.power, power=c.power,
preemphasis=c.preemphasis, preemphasis=c.preemphasis)
min_mel_freq=c.min_mel_freq,
max_mel_freq=c.max_mel_freq)
# Setup the dataset # Setup the dataset
train_dataset = Dataset( train_dataset = Dataset(

18
utils/audio.py
View File

@ -19,8 +19,6 @@ class AudioProcessor(object):
num_freq, num_freq,
power, power,
preemphasis, preemphasis,
min_mel_freq,
max_mel_freq,
griffin_lim_iters=None): griffin_lim_iters=None):
print(" > Setting up Audio Processor...") print(" > Setting up Audio Processor...")
@ -33,8 +31,6 @@ class AudioProcessor(object):
self.num_freq = num_freq self.num_freq = num_freq
self.power = power self.power = power
self.preemphasis = preemphasis self.preemphasis = preemphasis
self.min_mel_freq = min_mel_freq
self.max_mel_freq = max_mel_freq
self.griffin_lim_iters = griffin_lim_iters self.griffin_lim_iters = griffin_lim_iters
self.n_fft, self.hop_length, self.win_length = self._stft_parameters() self.n_fft, self.hop_length, self.win_length = self._stft_parameters()
if preemphasis == 0: if preemphasis == 0:
@ -54,7 +50,6 @@ class AudioProcessor(object):
n_fft = (self.num_freq - 1) * 2 n_fft = (self.num_freq - 1) * 2
return librosa.filters.mel( return librosa.filters.mel(
self.sample_rate, n_fft, n_mels=self.num_mels) self.sample_rate, n_fft, n_mels=self.num_mels)
# fmin=self.min_mel_freq, fmax=self.max_mel_freq)
def _normalize(self, S): def _normalize(self, S):
return np.clip((S - self.min_level_db) / -self.min_level_db, 0, 1) return np.clip((S - self.min_level_db) / -self.min_level_db, 0, 1)
@ -105,19 +100,6 @@ class AudioProcessor(object):
else: else:
return self._griffin_lim(S**self.power) return self._griffin_lim(S**self.power)
# def _griffin_lim(self, S):
# '''Applies Griffin-Lim's raw.
# '''
# S_best = copy.deepcopy(S)
# for i in range(self.griffin_lim_iters):
# S_t = self._istft(S_best)
# est = self._stft(S_t)
# phase = est / np.maximum(1e-8, np.abs(est))
# S_best = S * phase
# S_t = self._istft(S_best)
# y = np.real(S_t)
# return y
def _griffin_lim(self, S): def _griffin_lim(self, S):
angles = np.exp(2j * np.pi * np.random.rand(*S.shape)) angles = np.exp(2j * np.pi * np.random.rand(*S.shape))
S_complex = np.abs(S).astype(np.complex) S_complex = np.abs(S).astype(np.complex)