start

audio_processing.py

@@ -0,0 +1,98 @@
+import torch
+import numpy as np
+from scipy.signal import get_window
+import librosa.util as librosa_util
+
+
+def window_sumsquare(window,
+                     n_frames,
+                     hop_length=200,
+                     win_length=800,
+                     n_fft=800,
+                     dtype=np.float32,
+                     norm=None):
+    """
+    # from librosa 0.6
+    Compute the sum-square envelope of a window function at a given hop length.
+
+    This is used to estimate modulation effects induced by windowing
+    observations in short-time fourier transforms.
+
+    Parameters
+    ----------
+    window : string, tuple, number, callable, or list-like
+        Window specification, as in `get_window`
+
+    n_frames : int > 0
+        The number of analysis frames
+
+    hop_length : int > 0
+        The number of samples to advance between frames
+
+    win_length : [optional]
+        The length of the window function.  By default, this matches `n_fft`.
+
+    n_fft : int > 0
+        The length of each analysis frame.
+
+    dtype : np.dtype
+        The data type of the output
+
+    Returns
+    -------
+    wss : np.ndarray, shape=`(n_fft + hop_length * (n_frames - 1))`
+        The sum-squared envelope of the window function
+    """
+    if win_length is None:
+        win_length = n_fft
+
+    n = n_fft + hop_length * (n_frames - 1)
+    x = np.zeros(n, dtype=dtype)
+
+    # Compute the squared window at the desired length
+    win_sq = get_window(window, win_length, fftbins=True)
+    win_sq = librosa_util.normalize(win_sq, norm=norm)**2
+    win_sq = librosa_util.pad_center(win_sq, n_fft)
+
+    # Fill the envelope
+    for i in range(n_frames):
+        sample = i * hop_length
+        x[sample:min(n, sample+n_fft)] += win_sq[:max(0, min(n_fft, n - sample))]
+    return x
+
+
+def griffin_lim(magnitudes, stft_fn, n_iters=30):
+    """
+    PARAMS
+    ------
+    magnitudes: spectrogram magnitudes
+    stft_fn: STFT class with transform (STFT) and inverse (ISTFT) methods
+    """
+
+    angles = np.angle(np.exp(2j * np.pi * np.random.rand(*magnitudes.size())))
+    angles = angles.astype(np.float32)
+    angles = torch.autograd.Variable(torch.from_numpy(angles))
+    signal = stft_fn.inverse(magnitudes, angles).squeeze(1)
+
+    for i in range(n_iters):
+        _, angles = stft_fn.transform(signal)
+        signal = stft_fn.inverse(magnitudes, angles).squeeze(1)
+    return signal
+
+
+def dynamic_range_compression(x, C=1, clip_val=1e-5):
+    """
+    PARAMS
+    ------
+    C: compression factor
+    """
+    return torch.log(torch.clamp(x, min=clip_val) * C)
+
+
+def dynamic_range_decompression(x, C=1):
+    """
+    PARAMS
+    ------
+    C: compression factor used to compress
+    """
+    return torch.exp(x) / C

data_utils.py

@@ -0,0 +1,73 @@
+import random
+import numpy as np
+import hparams
+import torch
+import torch.utils.data
+import torch.nn.functional as F
+import os
+import pickle as pkl
+
+from text import text_to_sequence
+
+
+def load_filepaths_and_text(metadata, split="|"):
+    with open(metadata, encoding='utf-8') as f:
+        filepaths_and_text = [line.strip().split(split) for line in f]
+    return filepaths_and_text
+
+
+class TextMelLoader(torch.utils.data.Dataset):
+    def __init__(self, audiopaths_and_text, hparams):
+        self.audiopaths_and_text = load_filepaths_and_text(audiopaths_and_text)
+        random.seed(1234)
+        random.shuffle(self.audiopaths_and_text)
+
+    def get_mel_text_pair(self, audiopath_and_text):
+        # separate filename and text
+        file_name = audiopath_and_text[0][:10]
+        seq = f'{hparams.data_path}/preprocessed/sequence'
+        mel = f'{hparams.data_path}/preprocessed/melspectrogram'
+
+        with open(f'{seq}/{file_name}_sequence.pkl', 'rb') as f:
+            text = pkl.load(f)
+        with open(f'{mel}/{file_name}_melspectrogram.pkl', 'rb') as f:
+            mel = pkl.load(f)
+
+        return (text, mel)
+
+    def __getitem__(self, index):
+        return self.get_mel_text_pair(self.audiopaths_and_text[index])
+
+    def __len__(self):
+        return len(self.audiopaths_and_text)
+
+
+class TextMelCollate():
+    def __init__(self):
+        return
+
+    def __call__(self, batch):
+        # Right zero-pad all one-hot text sequences to max input length
+        input_lengths, ids_sorted_decreasing = torch.sort(
+            torch.LongTensor([len(x[0]) for x in batch]),
+            dim=0, descending=True)
+        max_input_len = input_lengths[0]
+
+        text_padded = torch.zeros(len(batch), max_input_len, dtype=torch.long)
+        for i in range(len(ids_sorted_decreasing)):
+            text = batch[ids_sorted_decreasing[i]][0]
+            text_padded[i, :text.size(0)] = text
+
+        # Right zero-pad
+        num_mels = batch[0][1].size(0)
+        max_target_len = max([x[1].size(1) for x in batch])
+
+        # include Spec padded and gate padded
+        mel_padded = torch.zeros(len(batch), num_mels, max_target_len)
+        output_lengths = torch.LongTensor(len(batch))
+        for i in range(len(ids_sorted_decreasing)):
+            mel = batch[ids_sorted_decreasing[i]][1]
+            mel_padded[i, :, :mel.size(1)] = mel
+            output_lengths[i] = mel.size(1)
+
+        return text_padded, input_lengths, F.pad(mel_padded, (5,10)), output_lengths