Transformer

HilbertTransformer

Extract instantaneous features from a time series via Hilbert transform.

Computes the analytic signal using the Hilbert transform and returns one of several derived representations: the amplitude envelope, instantaneous phase, instantaneous frequency, the quadrature component, or all three main features as a multi-column DataFrame.

Wraps scipy.signal.hilbert.

Quickstart

python

from sktime.transformations.hilbert import HilbertTransformer

estimator = HilbertTransformer(output_type='envelope', N=None, unwrap_phase=True, fs=1.0)

Parameters(4)

output_typestr, default=”envelope”: Which feature to extract. One of:
"envelope": instantaneous amplitude, |z(t)|
"phase": unwrapped instantaneous phase in radians
"frequency": instantaneous frequency in cycles per sample
"quadrature": imaginary part of the analytic signal
"all": envelope, phase and frequency as separate columns
Nint or None, default=None: Number of Fourier components (FFT length). If None, defaults to the length of the input.
unwrap_phasebool, default=True: Whether to unwrap the phase to remove 2-pi discontinuities.
fsfloat, default=1.0: Sampling frequency, used to scale instantaneous frequency into physical units (Hz) when output_type is "frequency" or "all".

Examples

>>> from sktime.transformations.hilbert import HilbertTransformer
>>> from sktime.datasets import load_airline
>>> y = load_airline ()
>>> t = HilbertTransformer (output_type = "envelope")
>>> y_envelope = t. fit_transform (y)