I am trying to minimise the noise in a spectrum of a variable. I have divided the time history in equal parts, I have multiplied every part of the time history by a Tukey window, done a FFT of each one of this window. After this procedure I have an adequate signal, I have a question: Should I compensate the energy lost with the Tukey window? and how?
Asked
Active
Viewed 143 times
2
-
Did you overlap the frames of windowed signal? See here: https://dsp.stackexchange.com/questions/13436/choosing-the-right-overlap-for-a-window-function – ZR Han Apr 29 '21 at 08:21
-
@ZRHan Yes. I overlap – Luca Mirtanini Apr 29 '21 at 08:27
-
If your window and noverlap satisfy COLA, the synthesized signal doesn’t lose any energy due to windowing. – ZR Han Apr 29 '21 at 08:37
-
How is it possible? It is always a mean of FT made on a "weakened" (due to the Tukey window) signal. – Luca Mirtanini Apr 29 '21 at 09:05
-
But you have overlapped and added the frames. – ZR Han Apr 29 '21 at 11:35
-
1I selected intervals of time history that overla, I did the fft of each one, and then the mean. what should I add? – Luca Mirtanini Apr 29 '21 at 12:17
-
Please see here: https://www.dsprelated.com/freebooks/sasp/Overlap_Add_OLA_STFT_Processing.html – ZR Han Apr 29 '21 at 13:20