def __init__(self, feature_type: FeatureType):
if not (FeatureType.MFCC == feature_type):
raise Exception("Only supports {}".format(FeatureType.MFCC.name))
self.data_pkl = get_dataset("savee_sr_22k_2sec_4-classes.pkl")
_data, pre_train_data = randomize_split(self.data_pkl, split_ratio=0.7)
rl_data, eval_data = randomize_split(_data, split_ratio=0.8)
self.data = rl_data
self.pre_train_data = pre_train_data
self.eval_data = eval_data
def get_data(self):
training_data, testing_data = randomize_split(self.data)
x_train_mfcc = np.array(
[d[FeatureType.MFCC.name] for d in training_data])
y_train_emo = np.array([d['y_emo'] for d in training_data])
y_train_gen = np.array([d['y_gen'] for d in training_data])
x_test_mfcc = np.array(
[d[FeatureType.MFCC.name] for d in testing_data])
y_test_emo = np.array([d['y_emo'] for d in testing_data])
y_test_gen = np.array([d['y_gen'] for d in testing_data])
return (x_train_mfcc, y_train_emo,
y_train_gen), (x_test_mfcc, y_test_emo, y_test_gen)
def __init__(self, feature_type: FeatureType) -> None:
self.data_pkl = get_dataset(
"signal-no-silent-4-class-dataset-2sec_sr_22k.pkl")
for d in self.data_pkl:
single_file = {}
feature = librosa.feature.mfcc(d['x'], sr=SR, n_mfcc=NUM_MFCC)
single_file[feature_type.name] = feature
single_file['y_emo'] = d['emo']
single_file['y_gen'] = d['gen']
self.data.append(single_file)
rl_data, pre_train_data = randomize_split(self.data, split_ratio=0.7)
self.data = rl_data
self.pre_train_data = pre_train_data
def __init__(self, sr: int = 44) -> None:
self.data_pkl = get_dataset('improv-4_Class-sr_' + str(sr) +
'k_2sec.pkl')
for d in self.data_pkl:
self.data.append({
FeatureType.MFCC.name: d['x'],
'y_emo': d['emo'],
'y_gen': d['gen'],
'path': d['path']
})
rl_data, pre_train_data = randomize_split(self.data, split_ratio=0.7)
self.data = rl_data
self.pre_train_data = pre_train_data