The time instants are roughly those in the center of each window used for the spectrum. The time between estimates is the difference between the length of the window and the overlap.
For example if you have N samples and a window of length L with an overlap of K, then the estimated time for the first window of L samples is (L/2)/Fs, the next is (L/2+(L-K))/Fs, the next is (L/2+2*(L-K))/Fs, etc., up to the last possible window of samples.