Assuming I understand waht you want to do, I would use the Signal Processing Toolbox buffer function to do that.
You would have to calculate ‘L’ (number of samples) based on your sampling frequency (in samples/time unit), so if the time units are seconds L = 0.1*sampling_frequency.
The buffer function returns the segments in column-major order, so you can do the fft on the output of your buffer call:
FT = buffer(your_signal, L)/L;
will return the scaled complex Fourier transform of each column.