Audio Precessing class, passing data fetching argummetns from config

utils/audio.py

@@ -1,108 +1,126 @@
+import os
 import librosa
+import pickle
 import numpy as np
 from scipy import signal
 
 _mel_basis = None
+global c
 
 
-def save_wav(wav, path):
-    wav *= 32767 / max(0.01, np.max(np.abs(wav)))
-    librosa.output.write_wav(path, wav.astype(np.int16), c.sample_rate)
+class AudioProcessor(object):
+
+    def __init__(self, sample_rate, num_mels, min_level_db, frame_shift_ms,
+                frame_length_ms, preemphasis, ref_level_db, num_freq, power,
+                griffin_lim_iters=None):
+        self.sample_rate = sample_rate
+        self.num_mels = num_mels
+        self.min_level_db = min_level_db
+        self.frame_shift_ms = frame_shift_ms
+        self.frame_length_ms = frame_length_ms
+        self.preemphasis = preemphasis
+        self.ref_level_db = ref_level_db
+        self.num_freq = num_freq
+        self.power = power
+        self.griffin_lim_iters = griffin_lim_iters
 
 
-def _linear_to_mel(spectrogram):
-    global _mel_basis
-    if _mel_basis is None:
-        _mel_basis = _build_mel_basis()
-    return np.dot(_mel_basis, spectrogram)
+    def save_wav(self, wav, path):
+        wav *= 32767 / max(0.01, np.max(np.abs(wav)))
+        librosa.output.write_wav(path, wav.astype(np.int16), self.sample_rate)
 
 
-def _build_mel_basis():
-    n_fft = (c.num_freq - 1) * 2
-    return librosa.filters.mel(c.sample_rate, n_fft, n_mels=c.num_mels)
+    def _linear_to_mel(self, spectrogram):
+        global _mel_basis
+        if _mel_basis is None:
+            _mel_basis = self._build_mel_basis()
+        return np.dot(_mel_basis, spectrogram)
 
 
-def _normalize(S):
-    return np.clip((S - c.min_level_db) / -c.min_level_db, 0, 1)
+    def _build_mel_basis(self):
+        n_fft = (self.num_freq - 1) * 2
+        return librosa.filters.mel(self.sample_rate, n_fft, n_mels=self.num_mels)
 
 
-def _denormalize(S):
-    return (np.clip(S, 0, 1) * -c.min_level_db) + c.min_level_db
+    def _normalize(self, S):
+        return np.clip((S - self.min_level_db) / -self.min_level_db, 0, 1)
 
 
-def _stft_parameters():
-    n_fft = (c.num_freq - 1) * 2
-    hop_length = int(c.frame_shift_ms / 1000 * c.sample_rate)
-    win_length = int(c.frame_length_ms / 1000 * c.sample_rate)
-    return n_fft, hop_length, win_length
+    def _denormalize(self, S):
+        return (np.clip(S, 0, 1) * -self.min_level_db) + self.min_level_db
 
 
-def _amp_to_db(x):
-    return 20 * np.log10(np.maximum(1e-5, x))
+    def _stft_parameters(self):
+        n_fft = (self.num_freq - 1) * 2
+        hop_length = int(self.frame_shift_ms / 1000 * self.sample_rate)
+        win_length = int(self.frame_length_ms / 1000 * self.sample_rate)
+        return n_fft, hop_length, win_length
 
 
-def _db_to_amp(x):
-    return np.power(10.0, x * 0.05)
+    def _amp_to_db(self, x):
+        return 20 * np.log10(np.maximum(1e-5, x))
 
 
-def preemphasis(x):
-    return signal.lfilter([1, -c.preemphasis], [1], x)
+    def _db_to_amp(self, x):
+        return np.power(10.0, x * 0.05)
 
 
-def inv_preemphasis(x):
-    return signal.lfilter([1], [1, -c.preemphasis], x)
+    def apply_preemphasis(self, x):
+        return signal.lfilter([1, -self.preemphasis], [1], x)
 
 
-def spectrogram(y):
-    D = _stft(preemphasis(y))
-    S = _amp_to_db(np.abs(D)) - c.ref_level_db
-    return _normalize(S)
+    def apply_inv_preemphasis(self, x):
+        return signal.lfilter([1], [1, -self.preemphasis], x)
 
 
-def inv_spectrogram(spectrogram):
-    '''Converts spectrogram to waveform using librosa'''
-
-    S = _denormalize(spectrogram)
-    S = _db_to_amp(S + c.ref_level_db)  # Convert back to linear
-
-    # Reconstruct phase
-    return inv_preemphasis(_griffin_lim(S ** c.power))
+    def spectrogram(self, y):
+        D = self._stft(self.apply_preemphasis(y))
+        S = self._amp_to_db(np.abs(D)) - self.ref_level_db
+        return self._normalize(S)
 
 
-def _griffin_lim(S):
-    '''librosa implementation of Griffin-Lim
-    Based on https://github.com/librosa/librosa/issues/434
-    '''
-    angles = np.exp(2j * np.pi * np.random.rand(*S.shape))
-    S_complex = np.abs(S).astype(np.complex)
-    y = _istft(S_complex * angles)
-    for i in range(c.griffin_lim_iters):
-        angles = np.exp(1j * np.angle(_stft(y)))
+    def inv_spectrogram(self, spectrogram):
+        '''Converts spectrogram to waveform using librosa'''
+        S = _denormalize(spectrogram)
+        S = _db_to_amp(S + self.ref_level_db)  # Convert back to linear
+        # Reconstruct phase
+        return inv_preemphasis(_griffin_lim(S ** self.power))
+
+
+    def _griffin_lim(self, S):
+        '''librosa implementation of Griffin-Lim
+        Based on https://github.com/librosa/librosa/issues/434
+        '''
+        angles = np.exp(2j * np.pi * np.random.rand(*S.shape))
+        S_complex = np.abs(S).astype(np.complex)
         y = _istft(S_complex * angles)
-    return y
+        for i in range(self.griffin_lim_iters):
+            angles = np.exp(1j * np.angle(_stft(y)))
+            y = _istft(S_complex * angles)
+        return y
 
 
-def _istft(y):
-    _, hop_length, win_length = _stft_parameters()
-    return librosa.istft(y, hop_length=hop_length, win_length=win_length)
+    def _istft(self, y):
+        _, hop_length, win_length = _stft_parameters()
+        return librosa.istft(y, hop_length=hop_length, win_length=win_length)
 
 
-def melspectrogram(y):
-    D = _stft(preemphasis(y))
-    S = _amp_to_db(_linear_to_mel(np.abs(D)))
-    return _normalize(S)
+    def melspectrogram(self, y):
+        D = self._stft(self.apply_preemphasis(y))
+        S = self._amp_to_db(self._linear_to_mel(np.abs(D)))
+        return self._normalize(S)
 
 
-def _stft(y):
-    n_fft, hop_length, win_length = _stft_parameters()
-    return librosa.stft(y=y, n_fft=n_fft, hop_length=hop_length, win_length=win_length)
+    def _stft(self, y):
+        n_fft, hop_length, win_length = self._stft_parameters()
+        return librosa.stft(y=y, n_fft=n_fft, hop_length=hop_length, win_length=win_length)
 
 
-def find_endpoint(wav, threshold_db=-40, min_silence_sec=0.8):
-    window_length = int(c.sample_rate * min_silence_sec)
-    hop_length = int(window_length / 4)
-    threshold = _db_to_amp(threshold_db)
-    for x in range(hop_length, len(wav) - window_length, hop_length):
-        if np.max(wav[x:x + window_length]) < threshold:
-            return x + hop_length
-    return len(wav)
+    def find_endpoint(self, wav, threshold_db=-40, min_silence_sec=0.8):
+        window_length = int(self.sample_rate * min_silence_sec)
+        hop_length = int(window_length / 4)
+        threshold = _db_to_amp(threshold_db)
+        for x in range(hop_length, len(wav) - window_length, hop_length):
+            if np.max(wav[x:x + window_length]) < threshold:
+                return x + hop_length
+        return len(wav)

utils/text/__init__.py

@@ -1,8 +1,8 @@
 #-*- coding: utf-8 -*-
 
 import re
-from Tacotron.utils.text import cleaners
-from Tacotron.utils.text.symbols import symbols
+from TTS.utils.text import cleaners
+from TTS.utils.text.symbols import symbols
 
 
 # Mappings from symbol to numeric ID and vice versa:

utils/text/symbols.py

@@ -7,7 +7,7 @@ Defines the set of symbols used in text input to the model.
 The default is a set of ASCII characters that works well for English or text that has been run
 through Unidecode. For other data, you can modify _characters. See TRAINING_DATA.md for details.
 '''
-from Tacotron.utils.text import cmudict
+from TTS.utils.text import cmudict
 
 _pad = '_'
 _eos = '~'