angles_spect = np.zeros((nRealizations,))
angles_clust = np.zeros((nRealizations,6))
for jj in range(nAngles,-1,-1): #range(nAngles + 1):
#Generate signal
rangle = np.arccos(1.0/2.0**0.25) + (np.pi/4.0)*(jj/(1.0*nAngles))
signal = np.zeros(n_sample_pnts)
#angles_true[jj] = rangle * 180 / np.pi
print('\n')
print('Creating signal...')
for i in range(n_sample_pnts):
(angles_true[jj],signal[i]) = sph.rand_sig(sample_pnts[i, :3].T, b, n_fibers, rangle)
s_energy = norm(signal,2)
snr = 20.0
tau = 10.0**(-snr/10.0) * s_energy
print('Signal to noise ratio: %0.5g' % snr)
#Start Monte Carlo simulations
for kk in range(nRealizations):
print('\n')
print('Realization: %2g' % kk)
# Make Rician noise
noiseR = np.random.randn(*signal.shape)
# Sample signal on lower degree quadrature
qsph1_16_132DP = np.loadtxt('data/qsph1-16-132DP.dat')
sample_pnts = qsph1_16_132DP[:, :3]
n_sample_pnts = 132
# Create reproducing-kernel (sparse representation) matrix
A_new = sph.interp_matrix(quad_pnts, sample_pnts, n_qpnts, n_sample_pnts, N)
# Create signal
print('Creating signal...')
n_fibers = 1 # number of Gaussian components (max n=3)
b = 4000 # s/mm^2
r_angle = -np.pi/4
signal = np.zeros(n_sample_pnts)
for i in range(n_sample_pnts):
signal[i] = sph.rand_sig(sample_pnts[i, :3].T, b, n_fibers, r_angle)[1]
SNR = []
nRealizations = 1
for kk in range(nRealizations):
# Make Rician noise
sigma = 0.0 # standard deviation
noiseR = sigma * np.random.randn(*signal.shape)
noiseI = sigma * np.random.randn(*signal.shape)
noise = noiseR + 1j*noiseI
SNR.append(10 * np.log10(norm(signal,2)/norm(noise,2)))
print('Signal to noise ratio: %0.5g' % SNR[kk])
# Add noise to signal
