help='The speekers gender list')
"""
This file is used to generate tfrecords for gender-sensitive PIT
speech seperation. Every tfrecords file contains:
inputs: [mix_speech_abs, max_speech_phase], shape:T*(fft_len*2)
labels: [spker1_speech_abs, apker2_speech_abs], shape:T*(fft_len*2)
gender: [spker1_gender, spker2_gender], shape:1*2
"""
args = parser.parse_args()
wavdir = args.wavdir
tfdir = args.tfdir
namelist = args.namelist
mkdir_p(tfdir)
if args.gender_list is not '':
apply_gender_info=True;
gender_dict = {}
fid = open(args.gender_list, 'r')
lines = fid.readlines()
fid.close()
for line in lines:
spk = line.strip('\n').split(' ')[0]
gender = line.strip('\n').split(' ')[1]
if gender.lower() == 'm':
gender_dict[spk] = 1;
else:
gender_dict[spk] = 0
def make_sequence_example(inputs, labels, genders):
parser.add_argument('wav_dir', help='The parent dit of mix,s1,s2')
parser.add_argument('list_dir', help='The parent dit of mix,s1,s2')
parser.add_argument('tfrecord_dir', help='TFRecords file dir')
parser.add_argument('--gender_list', default='', type=str, help='The speekers gender list')
parser.add_argument('--sample_rate', default=8000, type=int, help='sample rate of audio')
parser.add_argument('--window_size', default=256, type=int, help='window size for STFT')
parser.add_argument('--window_shift', default=128, type=int, help='window size for STFT')
args = parser.parse_args()
wav_dir = args.wav_dir
tfrecord_dir = args.tfrecord_dir
process_num = 8
list_dir = args.list_dir
mkdir_p(tfrecord_dir)
sample_rate = args.sample_rate
window_size = args.window_size
window_shift = args.window_shift
if args.gender_list is not '':
apply_gender_info = True
gender_dict = {}
fid = open(args.gender_list, 'r')
lines = fid.readlines()
fid.close()
for line in lines:
spk = line.strip('\n').split(' ')[0]
gender = line.strip('\n').split(' ')[1]
if gender.lower() == 'm':
gender_dict[spk] = 1