There is no single correct way of doing this. All windows will have their different spectral leakage profiles, and they're all useful in their own regards. The hann (hanning) window is often recommended for neuroimaging data, but so is rectwin (also known as boxcar), and dpss tapers.
Similarly, we can't tell you the best values for window length, noverlap, nfft, because we don't know the specifics of your data or what you intend to analyze.
I recommend watching this video tutorial for a good breakdown of time-frequency analysis. It uses the fieldtrip toolbox, but you can apply the knowledge towards the spectrogram() function as well.
