spectral_connectivity.transforms.Multitaper#

class Multitaper(time_series, sampling_frequency=1000, time_halfbandwidth_product=3, detrend_type='constant', time_window_duration=None, time_window_step=None, n_tapers=None, tapers=None, start_time=0, n_fft_samples=None, n_time_samples_per_window=None, n_time_samples_per_step=None, is_low_bias=True)[source]#

Bases: object

Transform time-domain signal(s) to the frequency domain by using multiple tapering windows.

Parameters:

time_series (array, shape (n_time_samples, n_trials, n_signals) or (n_time_samples, n_signals)) –
sampling_frequency (float, optional) – Number of samples per time unit the signal(s) are recorded at.
time_halfbandwidth_product (float, optional) – Specifies the time-frequency tradeoff of the tapers and also the number of tapers if n_tapers is not set.
detrend_type (string or None, optional) – Subtracting a constant or a linear trend from each time window. If None then no detrending is done.
start_time (float, optional) – Start time of time series.
time_window_duration (float, optional) – Duration of sliding window in which to compute the fft. Defaults to the entire time if not set.
time_window_step (float, optional) – Duration of time to skip when moving the window forward. By default, this equals the duration of the time window.
tapers (array, optional, shape (n_time_samples_per_window, n_tapers)) – Pass in a pre-computed set of tapers. If None, then the tapers are automically calculated based on the time_halfbandwidth_product, n_tapers, and n_time_samples_per_window.
n_tapers (int, optional) – Set the number of tapers. If None, the number of tapers is computed by 2 * time_halfbandwidth_product - 1.
n_time_samples_per_window (int, optional) – Number of samples in each sliding window. If time_window_duration is set, then this is calculated automically.
n_time_samples_per_step (int, optional) – Number of samples to skip when moving the window forward. If time_window_step is set, then this is calculated automically.
is_low_bias (bool, optional) – If True, excludes tapers with eigenvalues < 0.9

Attributes:

frequencies: Frequency of each frequency bin.
frequency_resolution: Range of frequencies the transform is able to resolve given the time-frequency tradeoff.
n_fft_samples: Number of frequency bins.
n_signals: Number of signals computed.
n_tapers: Number of desired tapers.
n_time_samples_per_step: If time_window_step is set, then calculate the n_time_samples_per_step based on the time window duration.
n_time_samples_per_window: Number of samples per time bin.
n_trials: Number of trials computed.
nyquist_frequency: Maximum resolvable frequency.
tapers: Returns the tapers used for the multitpaer function.
time: Time of each time bin.
time_window_duration: Duration of each time bin.
time_window_step: How much to each time window slides.

Methods

fft()

Compute the fast Fourier transform using the multitaper method.

Methods

fft

Compute the fast Fourier transform using the multitaper method.

Attributes

`frequencies`	Frequency of each frequency bin.
`frequency_resolution`	Range of frequencies the transform is able to resolve given the time-frequency tradeoff.
`n_fft_samples`	Number of frequency bins.
`n_signals`	Number of signals computed.
`n_tapers`	Number of desired tapers.
`n_time_samples_per_step`	If time_window_step is set, then calculate the n_time_samples_per_step based on the time window duration.
`n_time_samples_per_window`	Number of samples per time bin.
`n_trials`	Number of trials computed.
`nyquist_frequency`	Maximum resolvable frequency.
`tapers`	Returns the tapers used for the multitpaer function.
`time`	Time of each time bin.
`time_window_duration`	Duration of each time bin.
`time_window_step`	How much to each time window slides.

fft()[source]#

Compute the fast Fourier transform using the multitaper method.

Returns:: fourier_coefficients
Return type:: array, shape (n_time_windows, n_trials, n_tapers, n_fft_samples, n_signals)

property frequencies#: Frequency of each frequency bin.

property frequency_resolution#: Range of frequencies the transform is able to resolve given the time-frequency tradeoff.

property n_fft_samples#: Number of frequency bins.

property n_signals#: Number of signals computed.

property n_tapers#

Number of desired tapers.

Note that the number of tapers may be less than this number if the bias of the tapers is too high (eigenvalues > 0.9)

property n_time_samples_per_step#: If time_window_step is set, then calculate the n_time_samples_per_step based on the time window duration. If time_window_step and n_time_samples_per_step are both not set, default the window step size to the same size as the window.

property n_time_samples_per_window#: Number of samples per time bin.

property n_trials#: Number of trials computed.

property nyquist_frequency#: Maximum resolvable frequency.

property tapers#

Returns the tapers used for the multitpaer function. Tapers are the windowing function.

Returns:: tapers
Return type:: array_like, shape (n_time_samples_per_window, n_tapers)

property time#: Time of each time bin.

property time_window_duration#: Duration of each time bin.

property time_window_step#: How much to each time window slides.