Fourier on discrete but not sequential data

41 Views Asked by At

I have time series data, which is discrete as it is timestamped with microsecond resolution. It is not sequential, as in not every microsecond has a value. How would I go about Fourier in such a case? I figured I could just interpolate the missing data points, but this ended up with my machine running out of memory as the individual data points have quite some distance between them. I could also reduce the resolution, but before I do that I wanted to ask whether there is a better way than interpolation in the first place.

1

There are 1 best solutions below

0
On BEST ANSWER

There are libraries for nonequispaced fast Fourier transform or nonuniform fast Fourier transform. NFFT interpolates irregular data onto a regular grid followed by a standard FFT algorithm. If your data has some large gaps then I would try to break the data into chunks, and/or indeed reduce the resolution or decimate to study the low frequency spectra.