From 78603bd3c2374655a779ea179eef8be1cf84bb5c Mon Sep 17 00:00:00 2001
From: Matthew Scholefield <matthew331199@gmail.com>
Date: Wed, 5 Sep 2018 05:02:52 -0500
Subject: [PATCH] Add precise-add-noise script This is used to generate noise
 into a dataset

---
 precise/scripts/add_noise.py | 120 +++++++++++++++++++++++++++++++++++
 setup.py                     |   1 +
 2 files changed, 121 insertions(+)
 create mode 100755 precise/scripts/add_noise.py

diff --git a/precise/scripts/add_noise.py b/precise/scripts/add_noise.py
new file mode 100755
index 0000000..7965726
--- /dev/null
+++ b/precise/scripts/add_noise.py
@@ -0,0 +1,120 @@
+#!/usr/bin/env python3
+# Copyright 2018 Mycroft AI Inc.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+from itertools import chain
+from math import sqrt
+
+import numpy as np
+import os
+from glob import glob
+from os import makedirs
+from os.path import join, dirname, abspath
+from pip._vendor.distlib._backport import shutil
+from prettyparse import create_parser
+from random import random
+
+from precise.train_data import TrainData
+from precise.util import load_audio
+from precise.util import save_audio
+
+usage = '''
+    Create a duplicate dataset with added noise
+    
+    :folder str
+        Folder containing source dataset
+    
+    :-tg --tags-file str -
+        Tags file to optionally load from
+
+    :noise_folder str
+        Folder with wav files containing noise to be added
+
+    :output_folder str
+        Folder to write the duplicate generated dataset
+    ...
+'''
+
+
+class NoiseData:
+    def __init__(self, noise_folder: str):
+        self.noise_data = [
+            load_audio(file)
+            for file in glob(join(noise_folder, '*.wav'))
+        ]
+        self.noise_data_id = 0
+        self.noise_pos = 0
+        self.repeat_count = 0
+
+    def get_fresh_noise(self, n: int) -> np.ndarray:
+        noise_audio = np.empty(0)
+        while len(noise_audio) < n:
+            noise_source = self.noise_data[self.noise_data_id]
+            noise_chunk = noise_source[self.noise_pos:self.noise_pos + n - len(noise_audio)]
+            self.noise_pos += n - len(noise_audio)
+            if self.noise_pos >= len(noise_source):
+                self.noise_pos = 0
+                self.noise_data_id += 1
+                if self.noise_data_id >= len(self.noise_data):
+                    self.noise_data_id = 0
+                    self.repeat_count += 1
+                    if self.repeat_count == 100:
+                        print('Warning: Repeating noise 100+ times. Add more to prevent '
+                              'overfitting.')
+
+            noise_audio = np.concatenate([noise_audio, noise_chunk])
+        return noise_audio
+
+    def noised_audio(self, audio: np.ndarray) -> np.ndarray:
+        noise_data = self.get_fresh_noise(len(audio))
+        audio_volume = sqrt(sum(audio ** 2))
+        noise_volume = sqrt(sum(noise_data ** 2))
+        adjusted_noise = audio_volume * noise_data / noise_volume
+        ratio = 0.0 + 0.4 * random()
+        return ratio * adjusted_noise + (1.0 - ratio) * audio
+
+
+def main():
+    args = create_parser(usage).parse_args()
+    args.tags_file = abspath(args.tags_file)
+    args.folder = abspath(args.folder)
+    args.output_folder = abspath(args.output_folder)
+
+    data = TrainData.from_both(args.tags_file, args.folder, args.folder)
+    noise_data = NoiseData(args.noise_folder)
+    print('Data:', data)
+
+    def translate_filename(source: str) -> str:
+        assert source.startswith(args.folder)
+        relative_file = source[len(args.folder):].strip(os.path.sep)
+        return join(args.output_folder, relative_file)
+
+    all_filenames = sum(data.train_files + data.test_files, [])
+    for i, filename in enumerate(all_filenames):
+        print('{0:.2%}  \r'.format(i / (len(all_filenames) - 1)), end='', flush=True)
+
+        audio = load_audio(filename)
+        altered = noise_data.noised_audio(audio)
+        output_filename = translate_filename(filename)
+
+        makedirs(dirname(output_filename), exist_ok=True)
+        save_audio(output_filename, altered)
+
+    print('Done!')
+
+    if args.tags_file and args.tags_file.startswith(args.folder):
+        shutil.copy2(args.tags_file, translate_filename(args.tags_file))
+
+
+if __name__ == '__main__':
+    main()
diff --git a/setup.py b/setup.py
index c68de78..8138dc1 100644
--- a/setup.py
+++ b/setup.py
@@ -50,6 +50,7 @@ setup(
     ],
     entry_points={
         'console_scripts': [
+            'precise-add-noise=precise.scripts.add_noise:main',
             'precise-collect=precise.scripts.collect:main',
             'precise-convert=precise.scripts.convert:main',
             'precise-eval=precise.scripts.eval:main',