feat: analysis yodas ja000 for WADA SNR

.gitattributes

@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+*.json
+audio_analysis_results.json filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,176 @@
+# Created by https://www.toptal.com/developers/gitignore/api/python
+# Edit at https://www.toptal.com/developers/gitignore?templates=python
+
+### Python ###
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/#use-with-ide
+.pdm.toml
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/
+
+### Python Patch ###
+# Poetry local configuration file - https://python-poetry.org/docs/configuration/#local-configuration
+poetry.toml
+
+# ruff
+.ruff_cache/
+
+# LSP config files
+pyrightconfig.json
+
+# End of https://www.toptal.com/developers/gitignore/api/python

audio_analysis_results.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08645bedb6fd8fee832a0c14f74a1a376322dd6773b5e9452ff46451eba0837c
+size 206376891

audio_quality_analysis.py

@@ -0,0 +1,131 @@
+from datasets import load_dataset
+import librosa
+import IPython.display as ipd
+from IPython.display import Audio, display
+import random
+from concurrent.futures import ProcessPoolExecutor
+import numpy as np
+
+import json
+
+
+ds0 = load_dataset('espnet/yodas', 'ja000')
+print("finished loading ja000")
+
+def wada_snr(wav):
+    # Direct blind estimation of the SNR of a speech signal.
+    #
+    # Paper on WADA SNR:
+    #   http://www.cs.cmu.edu/~robust/Papers/KimSternIS08.pdf
+    #
+    # This function was adapted from this matlab code:
+    #   https://labrosa.ee.columbia.edu/projects/snreval/#9
+
+    # init
+    eps = 1e-10
+    # next 2 lines define a fancy curve derived from a gamma distribution -- see paper
+    db_vals = np.arange(-20, 101)
+    g_vals = np.array([0.40974774, 0.40986926, 0.40998566, 0.40969089, 0.40986186, 0.40999006, 0.41027138, 0.41052627, 0.41101024, 0.41143264, 0.41231718, 0.41337272, 0.41526426, 0.4178192 , 0.42077252, 0.42452799, 0.42918886, 0.43510373, 0.44234195, 0.45161485, 0.46221153, 0.47491647, 0.48883809, 0.50509236, 0.52353709, 0.54372088, 0.56532427, 0.58847532, 0.61346212, 0.63954496, 0.66750818, 0.69583724, 0.72454762, 0.75414799, 0.78323148, 0.81240985, 0.84219775, 0.87166406, 0.90030504, 0.92880418, 0.95655449, 0.9835349 , 1.01047155, 1.0362095 , 1.06136425, 1.08579312, 1.1094819 , 1.13277995, 1.15472826, 1.17627308, 1.19703503, 1.21671694, 1.23535898, 1.25364313, 1.27103891, 1.28718029, 1.30302865, 1.31839527, 1.33294817, 1.34700935, 1.3605727 , 1.37345513, 1.38577122, 1.39733504, 1.40856397, 1.41959619, 1.42983624, 1.43958467, 1.44902176, 1.45804831, 1.46669568, 1.47486938, 1.48269965, 1.49034339, 1.49748214, 1.50435106, 1.51076426, 1.51698915, 1.5229097 , 1.528578  , 1.53389835, 1.5391211 , 1.5439065 , 1.54858517, 1.55310776, 1.55744391, 1.56164927, 1.56566348, 1.56938671, 1.57307767, 1.57654764, 1.57980083, 1.58304129, 1.58602496, 1.58880681, 1.59162477, 1.5941969 , 1.59693155, 1.599446  , 1.60185011, 1.60408668, 1.60627134, 1.60826199, 1.61004547, 1.61192472, 1.61369656, 1.61534074, 1.61688905, 1.61838916, 1.61985374, 1.62135878, 1.62268119, 1.62390423, 1.62513143, 1.62632463, 1.6274027 , 1.62842767, 1.62945532, 1.6303307 , 1.63128026, 1.63204102])
+
+    # peak normalize, get magnitude, clip lower bound
+    wav = np.array(wav)
+    max_val = np.abs(wav).max()
+    if max_val == 0:
+        max_val = eps
+        
+    wav = wav / max_val
+    
+    abs_wav = np.abs(wav)
+    abs_wav[abs_wav < eps] = eps
+
+    # calcuate statistics
+    # E[|z|]
+    v1 = max(eps, abs_wav.mean())
+    # E[log|z|]
+    v2 = np.log(abs_wav).mean()
+    # log(E[|z|]) - E[log(|z|)]
+    v3 = np.log(v1) - v2
+
+    # table interpolation
+    wav_snr_idx = None
+    if any(g_vals < v3):
+        wav_snr_idx = np.where(g_vals < v3)[0].max()
+    # handle edge cases or interpolate
+    if wav_snr_idx is None:
+        wav_snr = db_vals[0]
+    elif wav_snr_idx == len(db_vals) - 1:
+        wav_snr = db_vals[-1]
+    else:
+        wav_snr = db_vals[wav_snr_idx] + \
+            (v3-g_vals[wav_snr_idx]) / (g_vals[wav_snr_idx+1] - \
+            g_vals[wav_snr_idx]) * (db_vals[wav_snr_idx+1] - db_vals[wav_snr_idx])
+
+    # Calculate SNR
+    dEng = sum(wav**2)
+    dFactor = 10**(wav_snr / 10)
+    dNoiseEng = dEng / (1 + dFactor) # Noise energy
+    dSigEng = dEng * dFactor / (1 + dFactor) # Signal energy
+    snr = 10 * np.log10(dSigEng / dNoiseEng)
+
+    return snr
+
+
+def preprocess_audio(data):
+    # �?ータが整数型�?�場合、浮動小数点型に変換
+    if data.dtype == np.int16:
+        data = data.astype(np.float32) / np.iinfo(np.int16).max
+    elif data.dtype == np.int32:
+        data = data.astype(np.float32) / np.iinfo(np.int32).max
+
+    # ス�?レオをモノラルに変換?���?要があれば?�?
+    if len(data.shape) == 2:
+        data = data.mean(axis=1)
+
+    return data
+
+# 音声データの前処理とSNR計算を行う関数
+def process_audio_data(item):
+    # 音声データの前処理
+    audio_data = item['audio']['array']
+
+    # 音声データが空でないことを確認
+    if len(audio_data) == 0:
+        return None
+
+    preprocessed_data = preprocess_audio(audio_data)
+    
+    # WADA-SNRを計算
+    snr = wada_snr(preprocessed_data)
+
+    # データからidを取得
+    uuid = item['utt_id']
+    transcription = item['text']
+
+    return {
+        "ファイル名": uuid,
+        "SNR値": snr,
+        "トランスクリプション": transcription
+    }
+
+import os
+
+if __name__ == '__main__':
+    ds = load_dataset('espnet/yodas', 'ja000', trust_remote_code=True)
+
+    print("データ数: ", ds['train'].dataset_size)
+
+    # CPUのコア数を取得
+    cpu_count = os.cpu_count()
+
+    # 並列�?��?で関数を実�?
+    with ProcessPoolExecutor(max_workers=cpu_count) as executor:
+        results = list(executor.map(process_audio_data, ds['train']))
+
+    # Noneを除去
+    results = [result for result in results if result is not None]
+
+    # 結果をJSONフ��イルに保存
+    with open('audio_analysis_results.json', 'w') as f:
+        json.dump(results, f, ensure_ascii=False, indent=4)
+
+    print("JSONファイルが保存されました")

generate_graph.py

@@ -0,0 +1,27 @@
+import json
+import numpy as np
+import matplotlib.pyplot as plt
+
+# 結果のJSONファイルを読み込む
+with open('audio_analysis_results.json', 'r') as f:
+    results = json.load(f)
+
+# SNR値のリストを取得
+snr_values = [result['SNR値'] for result in results]
+
+# ヒストグラムのパラメータを設定
+bins = np.linspace(-20, 100, 121)  # -20dBから60dBまで1dB刻みのビン
+plt.figure(figsize=(10, 6))
+plt.hist(snr_values, bins=bins, edgecolor='black', linewidth=1.2)
+plt.xlabel('WADA-SNR (dB)')
+plt.ylabel('Number of Samples')
+plt.title('Audio Quality Histogram')
+plt.xlim(-20, 100)
+plt.ylim(0, plt.ylim()[1] * 1.1)  # y軸の上限を10%拡張
+plt.grid(True)
+plt.tight_layout()
+
+# PNGファイルとして保存
+plt.savefig('audio_quality_histogram.png', dpi=300)
+
+print("ヒストグラムが audio_quality_histogram.png として保存されました。")