def load_dataset():
f1 = open(
'/misc/data15/reco/bhattgau/Rnn/Lists/spk_softmax/Train_feats_labs.plst'
)
lines = f1.readlines()
lines = [l.strip() for l in lines]
labelz = [int(l.split()[1]) for l in lines]
#labelz = labelz[:20]
features = [l.split()[0] for l in lines]
f2 = open(
'/misc/data15/reco/bhattgau/Rnn/Lists/spk_softmax/Valid_feats_labs.plst'
)
lines = f2.readlines()
lines = [l.strip() for l in lines]
val_labelz = [int(l.split()[1]) for l in lines]
#val_labelz = val_labelz[:20]
val_features = [l.split()[0] for l in lines]
n_samp = len(features)
maxlen = 800 #pad all utterances to this length
feat_dim = 20
nSpk = 98
dpth = '/misc/data15/reco/bhattgau/Rnn/Data/mfcc/Nobackup/VQ_VAD_HO_EPD/'
Data = np.zeros((n_samp, maxlen, feat_dim), dtype='float32')
Mask = np.zeros((n_samp, maxlen), dtype='float32')
#Targets = np.zeros((n_samp, nSpk), dtype='int32')
vn_samp = len(val_features)
val_Data = np.zeros((vn_samp, maxlen, feat_dim), dtype='float32')
val_Mask = np.zeros((vn_samp, maxlen), dtype='float32')
#Targets = np.zeros((n_samp, nSpk), dtype='int32')
for ind, f in enumerate(features):
fname = os.path.join(dpth, f + '.fea')
fi = htkmfc.HTKFeat_read(fname)
data = fi.getall()[:, :20]
Mask[ind, :data.shape[0]] = 1.0
pad = maxlen - data.shape[0]
data = np.vstack((data, np.zeros((pad, 20), dtype='float32')))
Data[ind, :, :] = data
for ind, f in enumerate(val_features):
fname = os.path.join(dpth, f + '.fea')
fi = htkmfc.HTKFeat_read(fname)
data = fi.getall()[:, :20]
val_Mask[ind, :data.shape[0]] = 1.0
pad = maxlen - data.shape[0]
data = np.vstack((data, np.zeros((pad, 20), dtype='float32')))
val_Data[ind, :, :] = data
return Data, Mask, np.asarray(
labelz, dtype='int32'), val_Data, val_Mask, np.asarray(val_labelz,
dtype='int32')
def main(argv):
parser = argparse.ArgumentParser()
parser.add_argument("input", help="input wav file")
parser.add_argument("output", help="output mfcc file")
args = parser.parse_args()
r = htkmfc.HTKFeat_read(args.input)
#print "Sample Period: {}".format(r.sampPeriod)
print vars(r)
data = r.getall()
np.savetxt(args.output, data, fmt='%6.2f')
def find_speech(scpname):
# reading scp
scplines = []
with open(scpname) as f:
for line in f:
fname = line.split("=")[1].split("[")[0]
label = line.split("=")[1].split("[")[0].split(
"/")[len(line.split("=")[1].split("[")[0].split("/")) -
1].split(".")[0]
beg = int(line.split("[")[1].split(",")[0])
end = int(line.split(",")[1].split("]")[0])
dur = end - beg
scplines.append([fname, label, beg, dur, end])
# reading htk
array_dict = {}
count = 0
for [fname, label, beg, dur, end] in scplines:
features = htkmfc.HTKFeat_read(fname)
array_dict[label + "_seg" + str(count)] = [
np.array(features.getall()[beg:end]), beg, dur, end
]
count += 1
return array_dict
for l in all_feats:
fbfeat = os.path.join(dpth, l)
spk_id = re.split('[/_.]', l)[2] + '_' + re.split('[/_.]', l)[4]
lab = tspk_phr.index(spk_id)
fname = re.split('[/.]', l)[2]
if os.path.exists(fbfeat):
LABELS.append(lab)
NAMES.append(fname)
mask = np.zeros((382, ), dtype='float32')
ff = htkmfc.HTKFeat_read(fbfeat)
data = ff.getall()
nframes = data.shape[0]
mask[:nframes] = 1.0
MASKS.append(mask)
padl = maxl - nframes
pad = np.zeros((padl, 40), dtype='float32')
datapad = np.vstack((data, pad))
nframes = datapad.shape[0]
ptr = 0
#give each frame a forward-backward context of 5 frames