-------
This method only returns values through the *ac* and *fr* parameters
kj = 2j*pi*self.freq_data.fftfreq()/self.csteerVecFormulation = steerVecTranslation(self.steer)
normFactor = self.sig_loss_norm()
for i in self.freq_data.indices:
if not fr[i]:
kji = kj[i, newaxis]
if self.r_diag:
// This case is not used at the moment (see Trait r_diag)
//==============================================================================
// One cannot use spectral decomposition when diagonal of csm is removed,
// as the resulting modified eigenvectors are not orthogonal to each other anymore.
// Therefor potentiating cannot be applied only to the eigenvalues.
// --> To avoid this the root of the csm (removed diag) is calculated directly.
// WATCH OUT: This doesn"t really produce good results.
//==============================================================================
csm = self.freq_data.csm[i]
fill_diagonal(csm, 0)
csmRoot = fractional_matrix_power(csm, 1.0 / self.gamma)
beamformerOutput, steerNorm = beamformerFreq(False, steerVecFormulation, False, 1.0, (self.r0, self.rm, kji, csmRoot[newaxis]))
beamformerOutput /= steerNorm // take normalized steering vec
// set (unphysical) negative output values to 0
After Change
def _get_digest( self ):
return digest( self )
def calc(self, ac, fr):
Calculates the Functional Beamformer result for the frequencies defined by :attr:`freq_data`
This is an internal helper function that is automatically called when
accessing the beamformer"s :attr:`~BeamformerBase.result` or calling
its :meth:`~BeamformerBase.synthetic` method.
Parameters
----------
ac : array of floats
This array of dimension ([number of frequencies]x[number of gridpoints])
is used as call-by-reference parameter and contains the calculated
value after calling this method.
fr : array of booleans
The entries of this [number of frequencies]-sized array are either
"True" (if the result for this frequency has already been calculated)
or "False" (for the frequencies where the result has yet to be calculated).
After the calculation at a certain frequency the value will be set
to "True"
Returns
-------
This method only returns values through the *ac* and *fr* parameters
i = self.freq_data.indicesself.steer_obj.f = (self.freq_data.fftfreq()[i]).tolist()
normFactor = self.sig_loss_norm()
for cntFreq in range(len(i)):
if not fr[i[cntFreq]]:
if self.r_diag:
// This case is not used at the moment (see Trait r_diag)
// It would need some testing as structural changes were not tested...
//==============================================================================
// One cannot use spectral decomposition when diagonal of csm is removed,
// as the resulting modified eigenvectors are not orthogonal to each other anymore.
// Therefor potentiating cannot be applied only to the eigenvalues.
// --> To avoid this the root of the csm (removed diag) is calculated directly.
// WATCH OUT: This doesn"t really produce good results.
//==============================================================================
csm = self.freq_data.csm[i[cntFreq]]
fill_diagonal(csm, 0)
csmRoot = fractional_matrix_power(csm, 1.0 / self.gamma)
beamformerOutput, steerNorm = self.steer_obj._beamformerCall(cntFreq, self.r_diag, 1.0, (csmRoot[newaxis],))
beamformerOutput /= steerNorm // take normalized steering vec
// set (unphysical) negative output values to 0
indNegSign = sign(beamformerOutput) < 0
beamformerOutput[indNegSign] = 0.0
else:
eva = array(self.freq_data.eva[i[cntFreq]][newaxis], dtype="float64") ** (1.0 / self.gamma)
eve = array(self.freq_data.eve[i[cntFreq]][newaxis], dtype="complex128")
beamformerOutput, steerNorm = self.steer_obj._beamformerCall(cntFreq, self.r_diag, 1.0, (eva, eve)) // takes all EigVal into account
beamformerOutput /= steerNorm // take normalized steering vec
ac[i[cntFreq]] = (beamformerOutput ** self.gamma) * steerNorm * normFactor // the normalization must be done outside the beamformer
fr[i[cntFreq]] = True
class BeamformerCapon( BeamformerBase ):
