def generate_npz(index, wav_files, dataset_path, preproc_data_path):
mfccs, phns = get_mfccs_and_phones(wav_files[index])
f_name = wav_files[index].replace(dataset_path, preproc_data_path).replace('WAV', 'npz')
if os.path.isfile(f_name):
return
else:
savez(f_name, mfccs=mfccs, phns=phns)
def generate_npz(wav_files, dataset_path, preproc_data_path):
print("Extracting and saving features from wav files...")
for i in tqdm(range(len(wav_files))):
mfccs, phns = get_mfccs_and_phones(wav_files[i])
f_name = wav_files[i].replace(dataset_path,
preproc_data_path).replace('WAV', 'npz')
if os.path.isdir(os.path.dirname(f_name)) == False:
os.makedirs(os.path.dirname(f_name))
savez(f_name, mfccs=mfccs, phns=phns)
def generate_npz(wav_files):
n_total = len(wav_files)
for i in range(n_total):
if (i % (n_total // 10)) == 0:
print((i / (n_total // 10)) * 10, "% is converted")
mfccs, phns = get_mfccs_and_phones(wav_files[i])
f_name = wav_files[i].replace('WAV', 'npz')
savez(f_name, mfccs=mfccs, phns=phns)
from data_load import get_mfccs_and_phones
import numpy as np
from hparam import hparam as hp
if __name__ == '__main__':
hp.set_hparam_yaml("TINIT2")
mfccs, phns = get_mfccs_and_phones(
"/home/cocoonmola/datasets/TIMIT2/TRAIN/DR3/FCMG0/SA1.WAV")
print(mfccs.shape)
print(phns.shape)
np.save("./mfccs", mfccs)
train_phns = None
test_mfccs = None
test_phns = None
if __name__ == "__main__":
dataset_path = "/home/john/datasets"
preproc_data_path = "./"
s = datetime.datetime.now()
train_wav_files = glob.glob(os.path.join(dataset_path, TIMIT_TRAIN_WAV))
test_wav_files = glob.glob(os.path.join(dataset_path, TIMIT_TEST_WAV))
print('Starting pre-processing train dataset...')
print("Extracting and saving features from wav files...")
for index in tqdm(range(len(train_wav_files))):
mfcc, phn = get_mfccs_and_phones(train_wav_files[index])
if index == 0:
train_mfccs = mfcc
train_phns = phn
else:
train_mfccs = np.concatenate((train_mfccs, mfcc))
train_phns = np.concatenate((train_phns, phn))
print('Pre-processing of train dataset has finished!')
print('Starting pre-processing test dataset...')
print("Extracting and saving features from wav files...")
for index in tqdm(range(len(test_wav_files))):
mfcc, phn = get_mfccs_and_phones(test_wav_files[index])
if index == 0: