Maybe something like the following?
- Link the full continuous recordings to the database (menu Review raw file)
- Epoch your data in regular blocks (a few seconds or a few minutes, depending on the time resolution you need), using the option "Split in time blocks" in the option window of the menu "Import to database"
- Select all the epochs in the Process1 box
- Select the process "Power spectrum density" or "FFT" to get a spectrogram for each block of data (you need to select the option "Save individual PSD values (for each trial)")
- Add the process "Standardize > Concatenate time" immediately after, run the pipeline
- In the new time-frequency file, the timing would not correspond to the original file: each time sample would correspond to the PSD of one block. You can edit the time vector manually in the file at the end if needed for display purposes.
I modified the concatenation process in order to create "timefreq" instead of "spectrum" file:
https://github.com/brainstorm-tools/brainstorm3/commit/c96ff6587826c718d26bbde5f5421d300cfa6e3f
Update Brainstorm before you try this.