Mercurial > hg > simscene-py

changeset 44:b7b1672b3c3b

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Reading and writing of files now is done by soundfile since there seems to be a bug with writing .wav files with librosa (mplayer would play them as rubbish). Added soundfile as a requirement.
author Emmanouil Theofanis Chourdakis <e.t.chourdakis@qmul.ac.uk>
date Mon, 09 Oct 2017 11:55:03 +0100
parents a8e3d63653b6
children 771dde08349a
files python/requirements.txt python/simscene.py
diffstat 2 files changed, 11 insertions(+), 7 deletions(-) [+]
line wrap: on
line diff
--- a/python/requirements.txt	Sun Oct 08 15:23:50 2017 +0100
+++ b/python/requirements.txt	Mon Oct 09 11:55:03 2017 +0100
@@ -4,3 +4,4 @@
 librosa
 matplotlib
 tabulate
+soundfile
--- a/python/simscene.py	Sun Oct 08 15:23:50 2017 +0100
+++ b/python/simscene.py	Mon Oct 09 11:55:03 2017 +0100
@@ -22,7 +22,10 @@
 # Librosa
 import librosa
 import librosa.display
-import librosa.output
+
+# PySoundfile
+
+import soundfile as sf
 
 # Matplotlib
 import matplotlib.pyplot as plt
@@ -73,7 +76,7 @@
             candidates = glob.glob('{}/event/{}*wav'.format(input_path, sampleid))
             chosen_fname = random.sample(candidates, 1)[0]
             
-            wav, SR = librosa.load(chosen_fname, sr=sr)
+            wav, SR = sf.read(chosen_fname)
         else:
             chosen_fname = random.sample(candidates, 1)[0]
 
@@ -365,7 +368,7 @@
             # If no patterns are found, search for normal audio files
             candidates = glob.glob('{}/background/{}*.wav'.format(input_path, backgrounds_df['sampleid'].loc[n]))           
             chosen_fname = random.sample(candidates, 1)[0]
-            wav, sr = librosa.load(chosen_fname, sr=SR)
+            wav, sr = sf.read(chosen_fname)
         else:
             chosen_fname = random.sample(candidates, 1)[0]
             wav, sr, pattern_timedict = render_pattern(chosen_fname, input_path)
@@ -541,7 +544,7 @@
             # If no patterns are found, search for normal audio files
             candidates = glob.glob('{}/event/{}*.wav'.format(input_path, events_df['sampleid'].loc[n]))           
             chosen_fname = random.sample(candidates, 1)[0]
-            wav, sr = librosa.load(chosen_fname, sr=SR)
+            wav, sr = sf.read(chosen_fname)
         else:
             chosen_fname = random.sample(candidates, 1)[0]
             wav, sr, pattern_timedict = render_pattern(chosen_fname, input_path)
@@ -685,7 +688,7 @@
             logging.debug('{}:New track ebr: {}db'.format(label, 20*np.log10(new_ebr)))
 
         if channel_mode == 'separate':
-            librosa.output.write_wav('{}/{}_event_track.wav'.format(output_path, label),
+            sf.write('{}/{}_event_track.wav'.format(output_path, label),
                                      track_arr/np.max(track_arr),
                                      SR)
             
@@ -786,9 +789,9 @@
         
     if channel_mode == 'mono':
         if append_to_filename:
-            librosa.output.write_wav('{}/scene_{}.wav'.format(output_path, append_to_filename), scene_arr, SR)
+            sf.write('{}/scene_{}.wav'.format(output_path, append_to_filename), scene_arr, SR)
         else:
-            librosa.output.write_wav('{}/scene.wav'.format(output_path), scene_arr, SR)
+            sf.write('{}/scene.wav'.format(output_path), scene_arr, SR)
 
     # Print timesdict