WSJ0数据集处理。原.wv数据集转换.wav方式
需要的工具：
- wsj0原数据集（LDC93S6A 或者 LDC93S6B）
- python3
- sph2pipe
python code:
FOR LDC93S6B:
"""
# example:
# 11-1.1/wsj0/si_tr_s/01t/01to030v.wv1 is converted to wav and
# stored in YOUR_PATH/wsj0/si_tr_s/01t/01to030v.wav
"""
import os
# the root dir for WSJ0 corpus
root_dir = "" #here add your original wsj0 dataset directory
# the disc number
disc_dir = []
for list_disc in os.listdir(root_dir):
if list_disc not in ["doc", "11-13.1"]: #doc file and 11-13.1 file do not contain .wv files
# the data dir for each disc
disc_dir.append(os.path.join(root_dir, list_disc, "wsj0"))
my_path = "YOUR PATH\\wsj0" #here add your target directory
if not os.path.exists(my_path):
os.mkdir(my_path)
# the sub_data dir for each disc
for i, list_sub_data in enumerate(disc_dir):
for sub_data_dir in os.listdir(list_sub_data):
s_dir = os.path.join(my_path, sub_data_dir)
if not os.path.exists(s_dir):
os.mkdir(s_dir)
if sub_data_dir[0][0] == 's':
datatype_dir = os.path.join(list_sub_data, sub_data_dir)
for list_spk in os.listdir(datatype_dir):
spk_dir = os.path.join(s_dir, list_spk)
spk_dir_abs = os.path.join(datatype_dir, list_spk)
if not os.path.exists(spk_dir):
os.mkdir(spk_dir)
for wv_file in os.listdir(spk_dir_abs):
if not wv_file.endswith('.wv1'): #only convert the .wv1 files, you can note the two lines to choose converting .wv1 and .wv2 files
continue
speech_dir = os.path.join(spk_dir_abs, wv_file)
target_name = wv_file.split(sep='.')[0] + '.wav'
target_dir = spk_dir + '\\' + target_name
# rif(wav)
cmd = "YOUR PATH\\sph2pipe -f rif " + speech_dir + " " + target_dir
os.system(cmd)
FOR LDC93S6A:
"""
# example:
# 11-1.1/wsj0/si_tr_s/01t/01to030v.wv1 is converted to wav and
# stored in YOUR_PATH/wsj0/si_tr_s/01t/01to030v.wav
"""
import os
# the root dir for WSJ0 corpus
root_dir = ""
# the disc number
disc_dir = []
for list_disc in os.listdir(root_dir):
if list_disc not in ["text", "11-13.1"]: #doc file and 11-13.1 file do not contain .wv files
# the data dir for each disc
disc_dir.append(os.path.join(root_dir, list_disc, "wsj0"))
my_path = "YOUR PATH\\wsj0"
if not os.path.exists(my_path):
os.mkdir(my_path)
# # the sub_data dir for each disc
for i, list_sub_data in enumerate(disc_dir):
for sub_data_dir in os.listdir(list_sub_data):
if (not sub_data_dir.startswith("si")) and (not sub_data_dir.startswith("sd")):
continue
s_dir = os.path.join(my_path, sub_data_dir)
if not os.path.exists(s_dir):
os.mkdir(s_dir)
if sub_data_dir[0][0] == 's':
datatype_dir = os.path.join(list_sub_data, sub_data_dir)
for list_spk in os.listdir(datatype_dir):
spk_dir = os.path.join(s_dir, list_spk)
spk_dir_abs = os.path.join(datatype_dir, list_spk)
if not os.path.exists(spk_dir):
os.mkdir(spk_dir)
for wv_file in os.listdir(spk_dir_abs):
if (not wv_file.endswith('.wv1')) and (not wv_file.endswith('.wv2')):
continue
speech_dir = os.path.join(spk_dir_abs, wv_file)
if wv_file.split('.')[1] == "wv1":
target_name = wv_file.split(sep='.')[0] + '.wav'
elif wv_file.split('.')[1] == 'wv2':
target_name = wv_file.split(sep='.')[0] + '_1.wav'
target_dir = spk_dir + '\\' + target_name
# rif == wav
cmd = "YOUR PATH\\sph2pipe -f wav " + speech_dir + " " + target_dir
os.system(cmd)
