def test_const(self):
x = np.array([0, 1, 2, 4, 7, 9, 10], dtype=float)
y = np.ones_like(x) * 3.0
for method in ['trapz', 'simps']:
self.assertAlmostEqual(3.0, uneven_mean(x, y, method=method))
self.assertAlmostEqual(3.0, uneven_mean(x + 20, y, method=method))
self.assertAlmostEqual(3.0, uneven_mean(x - 2, y, method=method))
self.assertAlmostEqual(-3.0, uneven_mean(x, -1 * y, method=method))
def calc_power_spectra(x, fft):
n_fft = fft.shape[1]
power_spectra = np.abs(fft) # new array
np.power(power_spectra, 2, out=power_spectra)
power_spectra = uneven_mean(x, power_spectra.real, axis=0)
power_spectra /= n_fft**2
return power_spectra
def calc_power_spectra(self, normalized_time: np.ndarray,
fft: np.ndarray) -> np.ndarray:
power_spectra = np.abs(fft) # new array
np.power(power_spectra, 2, out=power_spectra)
power_spectra = uneven_mean(normalized_time,
power_spectra.real,
axis=0)
# spectra_std = np.std(power_spectra.real, axis=0)
power_spectra /= self.n_fft**2
# spectra_std /= self.n_fft ** 2
return power_spectra
def calc_spectral_coherence(
self, normalized_time: np.ndarray, fft: Dict[Channel, np.ndarray],
power_spectra: Dict[Channel, np.ndarray]) -> np.ndarray:
scale = self.n_fft**2
fft1 = fft[self.channels[0]]
fft2 = fft[self.channels[1]]
power_spectra1 = power_spectra[self.channels[0]]
power_spectra2 = power_spectra[self.channels[1]]
cross_spectra = fft1 * fft2.conj()
cross_spectra_mean = uneven_mean(
normalized_time, cross_spectra, axis=0) / scale
return cross_spectra_mean / np.sqrt(power_spectra1 * power_spectra2)
def test_even_distribution(self):
n = 100000
x = np.arange(n)
y = np.random.randn(n)
self.assertAlmostEqual(uneven_mean(x, y), y.mean(), places=4)