def test_feature():
conf = load_yaml(datadir / "mlfb_vqvae.yml")
spkr_conf = load_yaml(datadir / "spkr.yml")
feat = Feature(datadir, conf["feature"], spkr_conf["SF1"], gl_flag=True)
feat.analyze(datadir / "SF1_10001.wav")
(datadir / "SF1_10001.h5").unlink()
(datadir / "SF1_10001_anasyn.wav").unlink()
def test_feature():
conf = load_yaml(ymlf)
spkr_conf = load_yaml(spkrymlf)
feat = Feature(datadir, conf["feature"], spkr_conf["SF1"])
feat.analyze(
datadir / "SF1_10001.wav",
synth_flag=True,
)
(datadir / "SF1_10001.h5").unlink()
(datadir / "SF1_10001_anasyn.wav").unlink()
def test_feature_8k():
conf = load_yaml(ymlf)
conf["feature"].update({
"fs": 8000,
"fftl": 256,
"fmin": 80,
"fmax": 3800,
"hop_size": 80,
"mlfb_dim": 80,
})
spkr_conf = load_yaml(datadir / "spkr.yml")
feat = Feature(datadir, conf["feature"], spkr_conf["SF1"])
feat.analyze(datadir / "SF1_10001_8k.wav", synth_flag=False)
(datadir / "SF1_10001_8k.h5").unlink()
def main():
dcp = "Extract aoucstic features"
parser = argparse.ArgumentParser(description=dcp)
parser.add_argument("--n_jobs", type=int, default=-1, help="# of CPUs")
parser.add_argument("--phase", type=str, default=None, help="phase")
parser.add_argument("--n_decode_samples",
type=int,
default=3,
help="# decode samples")
parser.add_argument("--conf",
type=str,
help="ymal file for network parameters")
parser.add_argument("--spkr_yml",
type=str,
help="yml file for speaker params")
parser.add_argument("--scpdir", type=str, help="scp directory")
parser.add_argument("--featdir", type=str, help="output feature directory")
args = parser.parse_args()
conf = load_yaml(args.conf)
spkr_conf = load_yaml(args.spkr_yml)
scp = open_scpdir(Path(args.scpdir) / args.phase)
featdir = Path(args.featdir) / conf["feature"]["label"] / args.phase
featsscp = featdir / "feats.scp"
if featsscp.exists():
featsscp.unlink()
for spkr in scp["spkrs"]:
logging.info("extract feature for {}".format(spkr))
wavs = [scp["wav"][uid] for uid in scp["spk2utt"][spkr]]
(featdir / spkr).mkdir(parents=True, exist_ok=True)
feat = Feature(featdir / spkr, conf["feature"], spkr_conf[spkr])
# create feats.scp
with open(featsscp, "a") as fp:
for uid in scp["spk2utt"][spkr]:
wavf = scp["wav"][uid]
h5f = str(featdir / spkr / (Path(wavf).stem + ".h5"))
fp.write("{} {}\n".format(uid, h5f))
# feature extraction with GliffinLim
Parallel(n_jobs=args.n_jobs)([
delayed(feat.analyze)(wavf, synth_flag=True)
for wavf in wavs[:args.n_decode_samples]
])
# feature extraction without GliffinLim
Parallel(n_jobs=args.n_jobs)([
delayed(feat.analyze)(wavf, synth_flag=False)
for wavf in wavs[args.n_decode_samples:]
])