def test_extract_feats_opensmile_config(recording):
config = OpenSmileConfig(
feature_set=opensmile.FeatureSet.ComParE_2016,
feature_level=opensmile.FeatureLevel.LowLevelDescriptors,
sampling_rate=recording.sampling_rate,
)
feature_extractor = OpenSmileExtractor(config=config)
y = feature_extractor.extract(recording.load_audio(),
recording.sampling_rate)
assert np.shape(y) == (1604, 65)
def test_extract_feats_config_defined_by_string(recording, feature_set,
feature_level, expected_shape):
config = OpenSmileConfig(
feature_set=feature_set,
feature_level=feature_level,
sampling_rate=recording.sampling_rate,
)
feature_extractor = OpenSmileExtractor(config=config)
y = feature_extractor.extract(recording.load_audio(),
recording.sampling_rate)
assert np.shape(y) == expected_shape
def test_num_frames(
feature_set,
feature_level,
):
sr = 8000
duration = 12.059
config = OpenSmileConfig(
feature_set=feature_set,
feature_level=feature_level,
sampling_rate=sr,
resample=True,
)
feature_extractor = OpenSmileExtractor(config=config)
num_frames = compute_num_frames(duration, feature_extractor.frame_shift,
sr)
num_samples = compute_num_samples(duration, sr)
signal = np.random.rand(1, num_samples)
y = feature_extractor.extract(signal, sr)
assert np.shape(y)[0] == num_frames
def test_extract_feats_default_config(recording):
feature_extractor = OpenSmileExtractor()
y = feature_extractor.extract(recording.load_audio(),
recording.sampling_rate)
assert np.shape(y) == (1604, 65)
def test_sampling_rate_mismatch(recording):
config = OpenSmileConfig(sampling_rate=8000)
feature_extractor = OpenSmileExtractor(config=config)
with pytest.raises(ValueError):
y = feature_extractor.extract(recording.load_audio(),
recording.sampling_rate)