emd.cycles.get_cycle_vector

emd.cycles.get_cycle_vector(phase, return_good=True, mask=None, imf=None, phase_step=4.71238898038469, phase_edge=0.2617993877991494, min_len=2)

Identify cycles within a instantaneous phase time-course.

Cycles are identified by large phase jumps and can optionally be tested to remove ‘bad’ cycles by criteria in Notes.

Parameters:

phasendarray

Input vector of Instantaneous Phase values

return_goodbool

Boolean indicating whether ‘bad’ cycles should be removed (Default value = True)

maskndarray

Vector of mask values that should be ignored (Default value = None)

imfndarray

Optional array of IMFs to used for control point identification when identifying good/bad cycles (Default value = None)

phase_stepscalar

Minimum value in the differential of the wrapped phase to identify a cycle transition (Default value = 1.5*np.pi)

phase_edgescalar

Maximum distance from 0 or 2pi for the first and last phase value in a good cycle. Only used when return_good is True (Default value = np.pi/12)

Returns:

ndarray

Vector of integers indexing the location of each cycle

Notes

Good cycles are those with

• A strictly positively increasing phase

• A phase starting within phase_step of zero (ie 0 < x < phase_edge)

• A phase ending within phase_step of 2pi (is 2pi-phase_edge < x < 2pi)

• A set of 4 unique control points (asc-zero, peak, desc-zero & trough)

Good cycles can be idenfied with:

>>> good_cycles = emd.cycles.get_cycle_vector(phase)

The total number of cycles is then

>>> good_cycles.max()

Indices where good cycles is zero do not contain a valid cycle

>>> bad_segments = good_cycles > -1

A single cycle can be isolated by matching its index, eg for the 5th cycle

>>> cycle_5_inds = good_cycles == 5