Tutorial 24: Time-frequency

Authors: Francois Tadel, Dimitrios Pantazis, Elizabeth Bock, Sylvain Baillet

This tutorial introduces how to compute time-frequency decomposition of MEG/EEG recordings and cortical currents using complex Morlet wavelets and Hilbert transforms.


Some of the MEG/EEG signal properties are difficult to access in time domain (graphs time/amplitude). A lot of the information of interest is carried by oscillations at certain frequencies, but the amplitude of these oscillations is sometimes a lot lower than the amplitude of the slower components of the signal, making them difficult to observe.

The averaging in time domain may also lead to a cancellation of these oscillations when they are not strictly locked in phase across trials. Averaging trials in time-frequency domain allows to extract the power of the oscillation regardless of the phase shifts. For a better understanding of this topic, we recommend the lecture of the following article: Bertrand O, Tallon-Baudry C (2000), Oscillatory gamma activity in humans: a possible role for object representation

In Brainstorm we offer two approaches for computing time-frequency decomposition (TF): the first one is based on the convolution of the signal with series of complex Morlet wavelets, the second filters the signal in different frequency bands and extracts the envelope of the filtered signals using the Hilbert transform.


Morlet wavelets

Complex Morlet wavelets are very popular in EEG/MEG data analysis for time-frequency decomposition. They have the shape of a sinusoid, weighted by a Gaussian kernel, and they can therefore capture local oscillatory components in the time series. An example of this wavelet is shown below, where the blue and red curves represent the real and imaginary part, respectively.

Contrary to the standard short-time Fourier transform, wavelets have variable resolution in time and frequency. For low frequencies, the frequency resolution is high but the time resolution is low. For high frenqucies, it's the opposite. When designing the wavelet, we basically decide a trade-off between temporal and spectral resolution.

To design the wavelet, we first need to choose a central frequency, ie. the frequency where we will define the mother wavelet. All other wavelets will be scaled and shifted versions of the mother wavelet. Unless interested in designing the wavelet at a particular frequency band, the default 1Hz should be fine.

Then, the desirable time resolution for the central frequency should be defined. For example, we may wish to have a temporal resolution of 3 seconds at frequency 1 Hz (default parameters). These two parameters, uniquely define the temporal and spectral resolution of the wavelet for all other frequencies, as shown in the plots below. Resolution is given in units of Full Width Half Maximum of the Gaussian kernel, both in time and frequency.

Edge effects

Users should pay attention to edge effects when applying wavelet analysis. Wavelet coefficients are computed by convolving the wavelet kernel with the time series. Similarly to any convolution of signals, there is zero padding at the edges of the time series and therefore the wavelet coefficients are weaker at the beginning and end of the time series.

From the figure above, which designs the Morlet wavelet, we can see that the default wavelet (central frequency fc=1Hz, FWHM_tc=3sec) has temporal resolution 0.6sec at 5Hz and 0.3sec at 10Hz. In such case, the edge effects are roughly half these times: 300ms in 5Hz and 150ms in 10Hz.

More precisely, if f is your frequency of interest, you can expect the edge effects to span over FWHM_t seconds: FWHM_t = FWHM_tc * fc / f / 2. Examples of such transients are given in the figures below.

edgeEffect5Hz.gif edgeEffect10Hz.gif

We also need to consider these edge effects when using the Hilbert transform approach. The band-pass filters used before extracting the signal envelope are relatively narrow and may cause long transients. To evaluate the duration of these edge effects for a given frequency band, use the interface of the process "Pre-process > Band-pass filter" or refer to the filters specifications (tutorial #10).


We will illustrate the time-frequency decomposition process with a simulated signal.

Process options

Comment: String that will be displayed in the database explorer to represent the output file.

Time definition

Frequency definition: Frequencies for which the power will be estimated at each time instant.

Morlet wavelet options

Compute the following measure:

Display: Time-frequency map

Display: Mouse and keyboard shortcuts

Mouse shortcuts

Keyboard shortcuts:

Figure popup menu

Display: Power spectrum and time series

Right-click on the file in the database or directly on the time-frequency figure to access these menus.

Normalized time-frequency maps

The brain is always active, the MEG/EEG recordings are never flat, some oscillations are always present in the signals. Therefore we are often more interested in the transient changes in the power at certain frequencies than at the actual power. A good way to observe these changes is to compute a deviation of the power with respect with a baseline.

There is another reason for which we are usually interested in standardizing the TF values. The power of the time-frequency coefficients are always lower in the higher frequencies than in the lower frequencies, the signal carries a lot less power in the faster oscillations than in the slow brain responses. This 1/f decrease in power is an observation that we already made with the power spectrum density in the filtering tutorial. If we represent the TF maps with a linear color scale, we will always see values close to zero in the higher frequency ranges. Normalizing each frequency separately with respect with a baseline helps obtaining more readable TF maps.

No normalization

The values we were looking at were already normalized (option: "1/f compensation" in the process options), but not with respect to a baseline. We will now compute the non-normalized power obtained with the Morlet wavelets and try the various options available for normalizing them.

Spectral flattening

Baseline normalization


Tuning the wavelet parameters

Time resolution

You can adjust the relative time and frequency resolution of your wavelet transformation by adjusting the parameters of the mother wavelet in the options of the process.

Examples for a constant central frequency of 1Hz with various time resolutions: 1.5s, 4s, 10s.

Frequency axis

You can also obtain very different representations of the data by changing the list of frequencies for which you estimate the power. You can change this in the options of the process.

Examples: Log 1:20:150, Log 1:300:150, Linear 15:0.1:25


Hilbert transform

We can repeat the same analysis with the other approach available for exploring the simulated signal in the time-frequency plane. The process "Frequency > Hilbert transform" first filters the signals in various frequency bands with a band-pass filter, then computes the Hilbert transform of the filtered signal. The magnitude of the Hilbert transform of a narrow-band signal is a measure of the envelope of this signal, and therefore gives an indication of the activity in this frequency band.

No normalization

Let's compute the same three results as before: non-normalized, 1/f compensation, baseline normalization.


Method specifications

MEG recordings: Single trials

Let's go back to our auditory oddball paradigm and apply the concepts to MEG recordings. We will use all the trials available for one condition to estimate the average time-frequency decomposition.

Spectral flattening

Baseline normalization

Things to avoid

Display: All channels

Three menus display the time-frequency of all the sensors with different spatial organizations. All the figures below represent the ERS/ERD-normalized average. They use the "jet" colormap, which is not the default configuration for these files. To get the same displays, change the colormap configuration:
right-click on the figure > Colormap: Stat2 > Colormap > jet.

Useful shortcuts for the first three figures:

Display: Topography

The menus below show the distribution of TF power over the sensors, for one time point and one frequency bin, very similarly to what was introduced in tutorial Visual exploration.

Useful shortcuts for these figures:



Similar calculations can be done at the level of the sources, either on the full cortex surface or on a limited number of regions of interests. We will start with the latter as it is usually an easier approach.


Full cortical maps

Computing the time-frequency decomposition for all the sources of the cortex surface is possible but complicated because it can easily generate gigantic files, completely out of the reach of most computers. For instance the full TF matrix for each trial we have here would be [Nsources x Ntimes x Nfrequencies] = [15000 x 361 x 40] double-complex = 3.2 Gb!

We have two ways of going around this issue: computing the TF decomposition for fewer frequency bins or frequency bands at a time, or as we did previously, use only limited number of regions of interest.


Unconstrained sources

In the current example, we are working with the simple case: sources with constrained orientations. The unconstrained case is more difficult to deal with, because we have to handle correctly the three orientations we have at each vertex.


Getting rid of the edge effects

To avoid making mistakes in the manipulation of the data and producing more readable figures, we encourage you to cut out the edge effects from your time frequency maps after computation.


On the hard drive

Right click one of the first TF file we computed > File > View file contents.

Structure of the time-frequency files: timefreq_*.mat

Useful functions


Additional documentation


Forum discussions

Feedback: Comments, bug reports, suggestions, questions
Email address (if you expect an answer):

Tutorials/TimeFrequency (last edited 2017-03-23 20:58:48 by ?MartinCousineau)