mne.time_frequency.tfr_multitaper(inst, freqs, n_cycles, time_bandwidth=4.0, use_fft=True, return_itc=True, decim=1, n_jobs=1, picks=None, average=True, verbose=None)[source]

Compute Time-Frequency Representation (TFR) using DPSS tapers.

inst : Epochs | Evoked

The epochs or evoked object.

freqs : ndarray, shape (n_freqs,)

The frequencies in Hz.

n_cycles : float | ndarray, shape (n_freqs,)

The number of cycles globally or for each frequency. The time-window length is thus T = n_cycles / freq.

time_bandwidth : float, (optional), defaults to 4.0 (3 good tapers).

Time x (Full) Bandwidth product. Should be >= 2.0. Choose this along with n_cycles to get desired frequency resolution. The number of good tapers (least leakage from far away frequencies) is chosen automatically based on this to floor(time_bandwidth - 1). E.g., With freq = 20 Hz and n_cycles = 10, we get time = 0.5 s. If time_bandwidth = 4., then frequency smoothing is (4 / time) = 8 Hz.

use_fft : bool, defaults to True

The fft based convolution or not.

return_itc : bool, defaults to True

Return inter-trial coherence (ITC) as well as averaged (or single-trial) power.

decim : int | slice, defaults to 1

To reduce memory usage, decimation factor after time-frequency decomposition. If int, returns tfr[…, ::decim]. If slice, returns tfr[…, decim].


Decimation may create aliasing artifacts.

n_jobs : int, defaults to 1

The number of jobs to run in parallel.

picks : array-like of int | None, defaults to None

The indices of the channels to decompose. If None, all available good data channels are decomposed.

average : bool, defaults to True

If True average across Epochs.

New in version 0.13.0.

verbose : bool, str, int, or None, defaults to None

If not None, override default verbose level (see mne.verbose() and Logging documentation for more).

power : AverageTFR | EpochsTFR

The averaged or single-trial power.

itc : AverageTFR | EpochsTFR

The inter-trial coherence (ITC). Only returned if return_itc is True.


New in version 0.9.0.