The term time-window or often just window is used to mean a short period of time during which something is possible, or during which you observe something, or during which you look for something.
Note that the paper introduces this concept earlier on when it says
For that, the recorded EMG-data were rectified and smoothed with a root mean square (RMS) function, working on a 250 ms time-window.
The full quote from the paper you mentioned says
Fig. 5 shows the CCF the healthy (left) and the spastic hemiparetic lower limb EMG signals of sTA and sGL over two windows of length N = 500.
In this case you can see from the graphs in Fig. 5 that the window-length is 500ms (0.5 seconds).
According to one online etymology dictionary, the term was first used with this meaning during the space-program to describe the short period of time during which a launch was possible, the launch window. And the English idiom window of opportunity came from this use. Note that in this context the window is defined by what's possible, or by some physical constraints that change over time, but in the paper you mentioned the window seems to be deliberately constrained by the experimenters, probably to get better accuracy after smoothing/filtering the signal, or to recognise patterns that happen within their defined window while excluding matching patterns that happen over longer periods.