def __init__(self, config=None):
super().__init__(config=config)
assert self.audio_featurizer.feat is not None
assert self.hparams.audio_config is not None
params_func = self.audio_featurizer.feat.params
params = params_func(self.hparams.audio_config)
energy_config = {
"type": "Framepow",
"window_length": params.window_length,
"frame_length": params.frame_length,
"remove_dc_offset": params.remove_dc_offset
}
self.energy_featurizer = AudioFeaturizer(energy_config)
pitch_config = {
"type": "Pitch",
"window_length": params.window_length,
"frame_length": params.frame_length,
"preEph_coeff": params.preEph_coeff,
"upper_frequency_limit": params.upper_frequency_limit
}
self.pitch_featurizer = AudioFeaturizer(pitch_config)
self.frame_length = params.frame_length
self.speakers_dict = {}
self.speakers_ids_dict = {}
self.text_featurizer = TextFeaturizer(self.hparams.text_config)
if self.hparams.data_csv is not None:
self.preprocess_data(self.hparams.data_csv)
def __init__(self, config=None):
super().__init__()
# hparams
self.hparams = register_and_parse_hparams(self.default_config,
config,
cls=self.__class__)
logging.info("hparams: {}".format(self.hparams))
self.audio_featurizer = AudioFeaturizer(self.hparams.audio_config)
self.feature_normalizer = FeatureNormalizer(self.hparams.cmvn_file)
def __init__(self, config=None):
super().__init__()
# hparams
self.hparams = register_and_parse_hparams(
self.default_config, config, cls=self.__class__)
logging.info("hparams: {}".format(self.hparams))
self.audio_featurizer = AudioFeaturizer(self.hparams.audio_config)
self.feature_normalizer = FeatureNormalizer(self.hparams.cmvn_file)
self.text_featurizer = TextFeaturizer(self.hparams.text_config)
if self.hparams.data_scps_dir is not None:
self.load_scps(self.hparams.data_scps_dir)
def __init__(self, config=None):
super().__init__()
# hparams
self.hparams = register_and_parse_hparams(self.default_config,
config,
cls=self.__class__)
logging.info("hparams: {}".format(self.hparams))
self.audio_featurizer = AudioFeaturizer(self.hparams.audio_config)
self.feature_normalizer = FeatureNormalizer(self.hparams.cmvn_file)
self.entries = []
self.speakers = []
if self.hparams.data_csv is not None:
self.load_csv(self.hparams.data_csv)
def __init__(self, config=None):
super().__init__(config=config)
self.speakers = []
self.audio_featurizer = AudioFeaturizer(self.hparams.audio_config)
self.spectrum_augmenter = SpecAugment(self.hparams.spec_augment_config)
self.feature_normalizer = FeatureNormalizer(self.hparams.cmvn_file)