def test_mask_single_channel(self):
shape = (3, 3)
signal = np.arange(np.prod(shape)).reshape(shape).astype(np.float32)
mask = lorenz_mask(signal, weight=1)
tc.assert_equal(
mask,
np.asarray([0, 0, 0, 0, 1, 1, 1, 1, 1],
dtype=np.float32).reshape(shape))
def test_mask_two_channels(self):
shape = (2, 3, 3)
signal = np.arange(np.prod(shape)).reshape(shape).astype(np.float32)
mask = lorenz_mask(signal, weight=1)
tc.assert_equal(
mask,
np.asarray([[[0, 0, 0], [0, 1, 1], [1, 1, 1]],
[[0, 0, 1], [1, 1, 1], [1, 1, 1]]],
dtype=np.float32))
def test_in_bound(self,
name,
sensor_axis=None,
lorenz_fraction=0.98,
weight=0.999):
mask = lorenz_mask(self.signal,
sensor_axis=sensor_axis,
lorenz_fraction=lorenz_fraction,
weight=weight)
assert_array_greater_equal(mask, 0.5 * (1 - weight))
assert_array_less_equal(mask, 0.5 * (1 + weight))
def test_shape(self,
name,
sensor_axis=None,
lorenz_fraction=0.98,
weight=0.999):
mask = lorenz_mask(self.signal,
sensor_axis=sensor_axis,
lorenz_fraction=lorenz_fraction,
weight=weight)
tc.assert_equal(
mask.shape,
tuple(v for i, v in enumerate([K, D, F, T]) if i != sensor_axis))
def beamform_mvdr_souden_with_lorenz_mask(
Y,
X_hat=None,
debug=False,
):
if X_hat is None:
X_hat = Y
X_mask = np.swapaxes(lorenz_mask(np.swapaxes(X_hat, -2, -1)), -2, -1)
N_mask = 1 - X_mask
return beamform_mvdr_souden_from_masks(
Y=Y,
X_mask=X_mask,
N_mask=N_mask,
debug=debug,
)
def test_multi_channel(self):
signal = self.signal[0][0]
mask1 = lorenz_mask(signal)
mask2 = lorenz_mask([signal, signal])
tc.assert_equal(mask1, mask2[0])
tc.assert_equal(mask1, mask2[1])
def test_weight_0(self, name, sensor_axis=None, lorenz_fraction=0.98):
mask = lorenz_mask(self.signal,
sensor_axis=sensor_axis,
lorenz_fraction=lorenz_fraction,
weight=0)
tc.assert_equal(mask, np.ones_like(mask) / 2)