1. I'm doing the overlap-add method for STFT and I quite understand the implementation if it's a 50% overlap (where I add the 2nd half of the previous frame to the 1st half of the current frame during synthesis).
    T1) [00,R1]
    T2)    [R1,R2] ---> output sum of T1(R1)+T2(R1)

For a 75% overlap, my frame hops for 1/4 of the window length. So if I have a L=128 window size, my next frame moves onto the next R=32 samples. Now for the synthesis, my question is how many previous frames should I add for the current frame. From how I understand it should be 4, so is the figure below correct?

    T1) [00,00,00,R1]
    T2)    [00,00,R1,R2]
    T3)       [00,R1,R2,R3]
    T4)          [R1,R2,R3,R4] ---> output sum of T1(R1)+T2(R1)+T3(R1)+T4(R1)

  1. And how does the windowing functions work for a 75% overlap? I thought most windows like Hanning/Hamming work properly only for 50% overlap. What's the advantage of a 75% overlap then?
  • 1
    $\begingroup$ There is no "Hanning". The guy's name was Von Hann and it's the Hann window. Take a look at this answer. All you need to make sure is that the windows add to 1. $\endgroup$ – robert bristow-johnson Feb 20 at 4:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.