def test_covariance_with_mask_independent_dim(self):
x = rand(2, 3, 4, data_type=np.complex128)
mask = np.random.uniform(0, 1, (
2,
4,
))
psd = get_power_spectral_density_matrix(x, mask)
tc.assert_equal(psd.shape, (2, 3, 3))
assert_positive_semidefinite(psd)
def generate_and_verify_psd(self, x_shape, mask_shape, psd_shape=None):
x, mask = self.generate_date(x_shape, mask_shape)
if mask_shape is None:
psd = get_power_spectral_density_matrix(x)
else:
psd = get_power_spectral_density_matrix(x, mask)
if psd_shape is not None:
tc.assert_equal(psd.shape, psd_shape)
assert_hermitian(psd)
assert_positive_semidefinite(psd)
def test_multiple_sources_for_source_separation(self):
x = rand(2, 3, 4, data_type=np.complex128)
mask = np.random.uniform(0, 1, (
5,
2,
4,
))
psd = get_power_spectral_density_matrix(x[np.newaxis, ...], mask)
tc.assert_equal(psd.shape, (5, 2, 3, 3))
assert_positive_semidefinite(psd)
def test_covariance_without_mask_independent_dim(self):
x = rand(1, 2, 3, 4, data_type=np.complex128)
psd = get_power_spectral_density_matrix(x)
tc.assert_equal(psd.shape, (1, 2, 3, 3))
assert_positive_semidefinite(psd)